STATUS: 200 OK (FOR NOW) 1 FIRE EXTINGUISHED TODAY

hi. I build the part of the application nobody sees until it breaks.

backend-focused developer working with APIs, databases, authentication, security, performance, and the occasional production fire.

"frontend gets the pixels. I get the logs."// fact
DEV // SHADES_MODE
Developer Vector Avatar - Cool Shades
STATUS: ONLINE// ready for backends
CORE STACK:Node • TS • Postgres • Redis
SPECIALTY:High-Concurrency APIs ⚡
LOCATION:India 🇮🇳 (Remote)
scroll ↓ the interesting stuff is underneath
ENGINEER PROFILE

about me (philosophy, skills & stack)

/* tech-agnostic // battle-tested // click any stack item for rationale */

Backend Visionary (3+ Years)

Building highly-resilient backend ecosystems with 3+ years of high-impact engineering. My code transforms complex requirements into scalable, AI-driven logic.

// MOD01READY

AI Frontier & Agentic LLMs

Deep exposure to the AI frontier: Orchestrating LLMs with Ollama, LMStudio, and Antigravity. I bridge the gap between human intent and machine execution.

// MOD02READY

Logical & High-Velocity Execution

Logical and results-driven full stack developer dedicated to building and optimizing user-focused applications with a calm and focused demeanor.

// MOD03READY

Uncompromising Quality

Warning: I am the candidate your recruiter warned you about. The good one.

// MOD04READY

TECHNICAL TOOLBELT & INFRASTRUCTURE MATRIX

⚡ CLICK ANY ITEM FOR USE CASES & RATIONALE (31 ITEMS)
👈 Click on any tech stack item above (e.g. HTML5, CSS3, JavaScript, TypeScript, React, Node.js, Fastify, PostgreSQL, Redis, Ollama, Kafka, AWS, Docker) to inspect real production use cases & engineering rationale!
COMPANY-FIRST HIERARCHYALL DELIVERABLES EXPANDED

engineering resume (experience & project post-mortems)

Structured by company experience. Each role details the exact systems architected, production challenges solved, and full code deliverables.

PRESENT // INDEPENDENT CONTRACTSMarch 31, 2026 – Present

Freelance & Independent Client Engineering

// Full Stack & Backend Specialist

Production Systems7 Apps
Ingestion Batch10k / 5s
Latency Delta< 5ms

"Architecting custom, production-grade enterprise backends, real-time collaboration engines, e-commerce platforms, and specialized B2B/B2C workflow systems for tech startups and international clients."

🚀 SYSTEMS ARCHITECTED UNDER FREELANCE & INDEPENDENT CLIENT ENGINEERING (7 PROJECTS):
Redis Distributed Lock (30s TTL)
concurrency

🛒 E-Commerce Platform

/* High-concurrency fashion e-commerce backend with multi-variant catalog, session-bound inventory reservation, and resilient checkout pipelines. */

Engineered a robust e-commerce engine designed to prevent race conditions during high-concurrency checkout waves while ensuring flawless media payload consistency across cart and payment steps.

🗺️ VISUAL DATA PIPELINE:
1.HTTP Checkout Request
2.Zod Schema Validation
3.Redis SETNX Inventory Lock (30s TTL)
4.PostgreSQL Stock Reservation Tx
5.Stripe Webhook Event Idempotency
⚙️ ARCHITECTURAL RESPONSIBILITIES & CODE DELIVERABLES:
  • Designed a multi-variant product catalog schema mapping complex SKU attributes (size, color, material, stock counts)
  • Implemented session-bound inventory reservation locks in Redis (`SETNX` with 30s TTL) to eliminate stock overselling during flash checkout rushes
  • Created a standardized `resolveImageUrl` media fallback utility ensuring nested variant thumbnails are cleanly resolved across instant 'Buy Now' and persistent cart flows
  • Engineered idempotent Stripe webhook handlers with event log tracking for automatic order fulfillment and automated inventory reconciliation
Stack: Next.js 16, Node.js, PostgreSQL, Prisma ORM, Stripe API, Zod schema validation, Redis locks.
10,000 Punch-ins -> 5s Batch Insert
concurrency

🏢 HRMS Lite (hrms-v1)

Human Resource Management System

/* Lightweight HRMS backend with biometric attendance ingestion, granular RBAC, automated payroll calculation, and instant JWT session revocation. */

Built to solve high-frequency morning rush bottlenecks when thousands of employees check in simultaneously via biometric devices, preventing database deadlock crashes while keeping employee data strictly governed.

🗺️ VISUAL DATA PIPELINE:
1.10,000+ Biometric Devices
2.Redis Ingestion List Queue
3.5-Second Cron Ingestion Worker
4.PostgreSQL Batch Insert (ON CONFLICT DO NOTHING)
5.Automated Payroll Ledger Update
⚙️ ARCHITECTURAL RESPONSIBILITIES & CODE DELIVERABLES:
  • Engineered a biometric punch-in ingestion queue using Redis lists, buffering 10,000+ morning 9:00 AM check-in events into 5-second transactional PostgreSQL batch inserts (`INSERT ... ON CONFLICT DO NOTHING`)
  • Implemented granular Role-Based Access Control (RBAC) middleware enforcing strict permission boundaries across Admin, HR Manager, and Employee roles
  • Designed automated payroll calculation engine evaluating tax deductions, leaves, unpaid mark-offs, and net monthly payouts
  • Built a JWT session revocation blacklist in Redis for instant employee offboarding and immediate credential invalidation
Stack: Fastify, TypeScript, PostgreSQL, Prisma, Redis in-memory queues, JWT authentication.
Async PDF Worker via BullMQ
concurrency

🧵 Garment Production & Invoicing Engine

/* Multi-currency B2B bulk invoicing engine, automated tax/GST routines, transactional garment production workflow tracker, and asynchronous PDF rendering workers. */

Designed for garment manufacturing factories to manage complex production cycles (cutting -> stitching -> QC -> packing) and generate multi-tier tax invoices without blocking server HTTP threads.

⚙️ ARCHITECTURAL RESPONSIBILITIES & CODE DELIVERABLES:
  • Engineered multi-currency B2B bulk invoicing engine with automated GST, regional tax, and volume discount calculation routines
  • Built transactional garment production workflow tracker tracing raw fabric rolls and trim materials through to finished unit inventory ledgers
  • Decoupled heavy 300DPI PDF document rendering from HTTP request handlers using Redis & BullMQ background job queues, streaming completed invoices via presigned download URLs
  • Created audit-logged inventory ledger maintaining historical cost snapshots for every fabric batch
Stack: Node.js, Express, PostgreSQL, Redis, BullMQ, PDFKit, Zod schemas.
Sub-Second Emergency Notification Push
realtime

🩸 BloodLink Emergency Matching Platform

/* Real-time emergency blood donation matching platform connecting seekers with nearby verified donors, donor health cooling state machine, and GIS radius search. */

A mission-critical life-saving application built for rapid emergency response, enforcing donor health safety while guaranteeing zero duplicate claims when emergency notifications trigger.

⚙️ ARCHITECTURAL RESPONSIBILITIES & CODE DELIVERABLES:
  • Built WebSocket real-time push notification service dispatching immediate emergency alerts to matching blood donors within a specified geographic radius
  • Implemented an automated 90-day donor cooling period state machine to enforce health safety and restrict re-donations before eligibility
  • Engineered Redis atomic locks (`SETNX`) on emergency donation requests to prevent duplicate acceptances when multiple donors respond simultaneously
  • Built an administrative verification dashboard for authenticating medical requests and managing user access
Stack: Next.js, Node.js, Fastify, Socket.io, MongoDB, Redis, Google Cloud VPS.
Ollama / GPU Node Fallback
ai security

🤖 AI Technical Interview Coach

/* Generative AI technical interview simulator with architect-level feedback, local LLM GPU fallback, and candidate communication analytics. */

An AI-powered interview simulator designed to conduct real-time technical and behavioral interview rounds, providing deep architect-level feedback on clarity, confidence, and system design.

⚙️ ARCHITECTURAL RESPONSIBILITIES & CODE DELIVERABLES:
  • Orchestrated Generative AI multi-agent workflows using Antigravity and Gemini Pro APIs for adaptive interview simulation
  • Integrated local LLM backup fallback via Ollama / LMStudio running on dedicated GPU nodes to ensure 100% operational uptime when cloud quotas exhaust
  • Built streaming response parser evaluating candidate answer clarity, technical depth, and communication nuances in real-time
  • Designed adaptive difficulty algorithms scaling question complexity dynamically based on candidate performance
Stack: Next.js 14, Antigravity, Ollama, Gemini Pro, AWS Lambda, TypeScript, PostgreSQL.
94% Payload Reduction via Vector Deltas
realtime

🎨 Collaborative Real-Time Whiteboard

/* Real-time room-based whiteboard application with Fastify WebSockets, Redis room state persistence, coordinate delta broadcasting, and canvas archives. */

A multiplayer canvas collaboration platform allowing isolated private rooms to draw, erase, chat, and store drawings without latency degradation.

⚙️ ARCHITECTURAL RESPONSIBILITIES & CODE DELIVERABLES:
  • Built Fastify WebSocket room server handling sub-5ms multi-user canvas drawing synchronization
  • Reduced WebSocket broadcast payload size by 94% by switching from full canvas frame snapshots to coordinate delta vector streams
  • Integrated Redis for in-memory room state storage, allowing instant canvas state recovery upon user page reload
  • Engineered long-term PostgreSQL snapshot archival service for preserving completed whiteboard sessions
Stack: Fastify, Next.js, WebSockets, Socket.io, Redis, PostgreSQL, Docker, PM2.
3-Min Single-Use Redis Nonce
ai security

📱 Secondary Device Linking & QR Auth

/* Multi-device authentication system using QR codes, 1-hour session timeout, 3-minute single-use Redis nonces, and CryptoJS encryption. */

Designed to provide seamless multi-device linking for banking and secure enterprise applications by scanning single-use QR codes from primary mobile sessions.

⚙️ ARCHITECTURAL RESPONSIBILITIES & CODE DELIVERABLES:
  • Designed multi-device authentication protocol allowing users to securely pair up to 2 secondary devices per active session
  • Implemented short-lived 3-minute QR code validity powered by dynamic single-use Redis TTL nonces to prevent replay attacks
  • Built CryptoJS payload encryption for secure device handshake exchanges over public networks
  • Enforced strict 1-hour session expiration and automatic token revocation
Stack: Next.js, Node.js, PostgreSQL, CryptoJS, JWT, Redis, Docker, PM2.
CORE EMPLOYMENT // FINTECHUntil March 31, 2026

Zeksta Technology Pvt Ltd

// Software Engineer (Fintech Lead)

Code Delivery50k+ Lines / 50 Days
API Modules4 Core / 45+ REST APIs
Commit Cadence173 Commits

"Lead Backend Engineer for the Sangamam Cooperative Banking Ecosystem. Single-handedly architected and delivered core financial modules under crushing time constraints with extreme velocity."

🚀 SYSTEMS ARCHITECTED UNDER ZEKSTA TECHNOLOGY PVT LTD (1 PROJECTS):
Two-Phase Commit ACID Integrity
fintech

🏦 Sangamam Core Banking Platform

Customer & Agent Mobile Backend Ecosystem

/* Architected and delivered 4 core banking modules from scratch, shipping 20+ major features, 45+ REST APIs, and 50,000+ lines of code within 3 months. */

Demonstrated high-velocity execution by maintaining a relentless 173-commit cadence across 50 active engineering days under crushing deadlines. Built the entire backend infrastructure powering mobile customer onboarding, agent collection apps, fund transfers, and administrative reporting.

🗺️ VISUAL DATA PIPELINE:
1.Mobile App MPIN & Device Binding
2.Redis Rate-Limit & OTP Nonce Check
3.PostgreSQL Two-Phase Commit Transaction
4.Atomic Debit + Credit Balance Execution
5.Double-Entry General Ledger Posting
⚙️ ARCHITECTURAL RESPONSIBILITIES & CODE DELIVERABLES:
  • ⚙️ Mobile Onboarding & Device Security: Designed and shipped customer and agent mobile app backends with Aadhaar OTP verification, MPIN setup, account linking, and strict device binding for maximum security.
  • 💸 Fund Transfer Engine: Built intra-bank transfer flow utilizing a two-phase commit pattern in PostgreSQL transactions, atomic debit+credit execution, Redis rate-limiting, counterparty visibility, and automated commission calculations.
  • 📊 Admin Reporting & Accounting Engine: Built a comprehensive reporting engine from scratch (4,000+ lines of logic) serving as the data backbone for admin dashboards with granular branch-level RBAC.
  • 📒 General Ledger System: Engineered full double-entry General Ledger (GL) voucher accounting system requiring multi-step balancing workflows and document attachments.
  • ☁️ Cloud Infrastructure & AWS ECS: Orchestrated resilient containerized microservices on AWS ECS, utilizing S3 for compliance document storage, behind secure API gateway load balancers.
  • 🛠️ Standards & Documentation: Established baseline database migrations, standardized TypeScript types, and 100% Swagger API documentation adopted across the entire platform scale.
Stack: Node.js, Express, Sequelize ORM, PostgreSQL, Redis, AWS ECS, AWS S3, Docker, PM2, TypeScript, Swagger.
ENTERPRISE & GOVT E-SERVICESPrevious Role

Quantela & Innovation Lab

// Associate Software Engineer

Govt ModulesCivil Court & Land
Record Scale1M+ Active Records
Dev Search AI+60% Speedup

"Engineered mission-critical backend modules for major government civil court web applications (eNibandan) and land transaction systems (MPWebGIS)."

🚀 SYSTEMS ARCHITECTED UNDER QUANTELA & INNOVATION LAB (3 PROJECTS):
Automated 30-Day Objection Scheduler
govt

📜 Civil Court Marriage Registration Engine

eNibandan Govt Civil Court Web Module

/* Government civil court web application module handling the full lifecycle of legal marriage applications, multi-tier officer approvals, and automated 30-day notice objection periods. */

Active lead backend engineer on the second largest civil court module (following land registration), handling complete workflow automation, notice period scheduling, and biometric eKYC verification.

🗺️ VISUAL DATA PIPELINE:
1.Citizen Online Form Submission
2.Assistant Officer Document Verification
3.Sub-Registrar (SRO) Approval
4.Automated 30-Day Objection Cron Scheduler
5.OSR Biometric eKYC Certificate Release
👥 MULTI-ROLE APPROVAL WORKFLOW:
  • Citizen Role: Bride and bridegroom details registration, secure application payment processing.
  • Assistant to Sub-Registrar: Document verification, authority to approve or send back for corrections.
  • Sub-Registrar (SRO): Comprehensive review, approval/rejection authority, entering approved applications into mandatory 30-day notice objection period.
  • Operator to Sub-Registrar (OSR): Biometric verification (eKYC) and final marriage certificate record issuance.
⚙️ ARCHITECTURAL RESPONSIBILITIES & CODE DELIVERABLES:
  • Engineered complete 4-role state machine managing transitions between Citizen -> Assistant -> Sub-Registrar -> OSR eKYC
  • Automated 30-day objection notice tracking using node-cron schedulers, automatically transitioning un-objected applications to appointment-ready status
  • Integrated secure biometric eKYC verification data handling and digital certificate record generation
  • Optimized PostgreSQL database schemas for legal compliance and auditability
Stack: Node.js, Express, PostgreSQL, Knex.js, Cron schedulers, eKYC integration.
48h Auto-Release & GIS Partition Logic
govt

MAP MP Land Transaction Premutation System

MPWebGIS / IGRS Department Module

/* MP Government land transaction system facilitating land partition sales, GIS visual selection, real-time availability validation, and automated payment deadline releases. */

Facilitates seller and buyer land partition transactions by integrating GIS visual land sketch selection with real-time double-transaction prevention.

⚙️ ARCHITECTURAL RESPONSIBILITIES & CODE DELIVERABLES:
  • Developed and maintained the Premutation Module verifying land sketch transactions submitted to the IGRS department
  • Integrated GIS-based visual land selection enabling users to select partition boundaries directly via GIS interfaces
  • Integrated real-time availability checks preventing concurrent transaction attempts on the same land partition
  • Enforced 48-hour payment validation rules: if payment is incomplete after 48h, cron automatically releases the land back to the public pool; if left un-submitted for 72h, it is auto-rejected
  • Optimized high-volume land record database queries using Knex.js and Redis temporary caching, handling peak loads of 1M+ active records
Stack: Node.js, Express.js, Knex.js, PostgreSQL, Redis caching, Cron schedulers, GIS APIs.
-60% Search Lead Time / Terraform IaC
ai security

🧠 Enterprise AI Search & IaC Infrastructure

/* Local LLM integration with Ollama and Terraform cloud automation for enterprise microservices. */

Integrated local LLMs (Ollama) into internal developer tools, reducing search time by 60% and automating infrastructure deployment across cloud environments.

⚙️ ARCHITECTURAL RESPONSIBILITIES & CODE DELIVERABLES:
  • Integrated local LLM runtime (Ollama) for internal code and documentation search, speeding up dev search by 60%
  • Automated infrastructure provisioning across AWS using Terraform, reducing deployment lead time significantly
  • Optimized PostgreSQL query layer for 40% improvement in AWS cloud resource utilization
Stack: Node.js, Fastify, Ollama, Terraform, AWS, PostgreSQL.
Architect Vector Avatar
ENGINEERING MENTAL MODEL

how I think (principles from the trenches)

/* 9 rules I live by when building backends */

1. Build the API first

// contract > UI

Decouple backend contract design from UI implementation. A clean API doesn't care if the caller is React, a mobile app, or a curl command.

RULE #1ENFORCED ✅

2. Understand data before writing queries

// EXPLAIN ANALYZE always

Draw the ERD diagram and query execution plan before writing complex joins. Indexes exist because databases also get tired.

RULE #2ENFORCED ✅

3. Validate input at the boundary

// input = suspect until proven valid

Never trust frontend validation. The client is an untrusted remote CLI. Validate schemas with strict Zod parsing at the API gateway.

RULE #3ENFORCED ✅

4. Assume users will send weird requests

// expect the unexpected

If your endpoint accepts a string, someone will send a 40MB PDF or a SQL payload. Handle edge cases defensively with explicit limits.

RULE #4ENFORCED ✅

5. Authentication is not authorization

// roles are not vibes

Knowing WHO someone is (Authentication ✅) does not mean they get to touch or delete the requested resource (Authorization ❌).

RULE #5ENFORCED ✅

6. Logs are part of the product

// debug-ready logs

Logging 'Error: request failed' is useless. Log structured context: correlation ID, user ID, payload digest, latency, and stack trace.

RULE #6ENFORCED ✅

7. Performance must be measured, not guessed

// p99 > average latency

Don't guess where the latency bottleneck is. Run JMeter stress suites, inspect p99 distributions, and profile socket queues.

RULE #7ENFORCED ✅

8. Security is an architectural constraint

// zero trust by default

Security is not a checkbox you review before launch. Rate limits, CORS, TLS, least privilege, and parameter sanitization belong in core architecture.

RULE #8ENFORCED ✅

9. Production is the final exam

// local host = optimism

Your system works on your machine? Great. Production is where real concurrency, network drops, and unexpected edge cases evaluate your code.

RULE #9ENFORCED ✅
MAJOR ENGINEERING SECTION

How Requirements Become Code (the engineering pipeline)

Good developers do not immediately start writing controllers. Here is how feature requests are analyzed, modeled, secured, and implemented for production systems.

/* requirements → architecture → code */

END-TO-END ENGINEERING PIPELINE

1. Business Requirement
2. Clarify Requirement
3. Functional Requirements
4. Non-Functional Requirements
5. Constraints
6. Domain Model
7. Architecture
8. API Contract
9. Database Design
10. Security Rules
11. Implementation
12. Testing
13. Observability
14. Deployment
15. Feedback
16. Iteration
"The requirement said ‘add a button’. Somehow we ended up discussing transactions, RBAC and database indexes."
END-TO-END CASE STUDY

Money Transfer Architecture (₹10,000 Case Study)

Tracing a real financial transaction through business rules, DB row locks, idempotency, security, outbox events & observability.

"Customer transfers ₹10,000" → Banking Grade Pipeline
01. REQUIREMENT & BUSINESS RULES

Business Statement: "Customer transfers ₹10,000 to Beneficiary"

RULES
Rule 1: Source account status must be ACTIVE & verified via MFA session token
Rule 2: Available account balance >= ₹10,000 (after reserving pending holds)
Rule 3: Daily transaction limit check (₹10,000 + today's total <= ₹50,000 limit)
Rule 4: Beneficiary account must exist, be ACTIVE, and be unblocked
Rule 5: Request must be authenticated with customer JWT and signed payload
// business rules must be validated before touching money balances
INTERACTIVE STEPPER

Real Requirement Transformation

Tracing a simple business sentence through all 10 engineering stages.

// requirement → engineering decision → code
#01

Business Requirement

What the stakeholder said
INPUT
"Customers should be able to place an order online."
// simple business sentence. zero technical details yet.
PRE-IMPLEMENTATION ANALYSIS

Before I Touch the Keyboard

Essential questions clarified before writing a single line of API code.

"Most bugs are born before the code exists."
QUESTION #1Identity

"Who is the user?"

Customer, Admin, Partner API client, or internal background job?

QUESTION #2Scope

"What problem are we solving?"

Are we solving a real operational pain point or building an unnecessary abstraction?

QUESTION #3Happy Path

"What is the expected behavior?"

What does success look like, and what data payload should be returned?

QUESTION #4Failure

"What happens when it fails?"

Do we return 4xx/5xx errors, trigger retries, or fail gracefully with fallbacks?

QUESTION #5Idempotency

"What happens when the same request arrives twice?"

Will duplicate POST requests create duplicate orders or be safely deduplicated via Idempotency-Key?

QUESTION #6RBAC

"Who is allowed to perform the action?"

What role bitmask or policy is required to access this endpoint?

QUESTION #7Validation

"What data is required?"

What fields are required in the payload, and what are their strict Zod constraints?

QUESTION #8Privacy

"What data should never be collected?"

Are we accidentally storing raw credit cards, unhashed passwords, or PII?

QUESTION #9Scale

"What are the expected traffic levels?"

Is this 10 requests per hour or a 5,000 req/sec flash sale?

QUESTION #10Threading

"What happens under concurrency?"

Will simultaneous requests cause SQL race conditions on stock or account balances?

QUESTION #11Compliance

"What must be audited?"

Do financial, security, or data mutations need immutable audit log records?

QUESTION #12Transactions

"What must be reversible?"

Can orders be canceled, payments refunded, or inventory reservations released?

QUESTION #13Integrations

"What are the external dependencies?"

Are we relying on Stripe, Twilio, SendGrid, or third-party webhooks?

QUESTION #14Legal

"What are the regulatory constraints?"

Does GDPR, PCI-DSS, or SOC2 compliance dictate data retention policies?

QUESTION #15Resilience

"What happens if a third-party service is unavailable?"

Do we have circuit breakers, fallback queues, or timeouts in place?

REQUIREMENT TAXONOMY

Functional vs Non-Functional Requirements

"The endpoint working is not the whole requirement."
WHAT THE SYSTEM DOESFUNCTIONAL

Core features, business capabilities, user inputs, and output behaviors.

  • Create customer order and record cart items
  • Update user profile & delivery addresses
  • Process payment charge via credit card gateway
  • Generate monthly PDF invoice reports
HOW WELL THE SYSTEM BEHAVESNON-FUNCTIONAL

System qualities, performance limits, security boundaries, and operational constraints.

  • Response Time: p99 latency < 200ms
  • Availability: 99.9% uptime SLA guarantees
  • Security: OWASP Top 10 + RBAC bitmask enforcement
  • Auditability: Immutable audit logs for state changes
PRECISION SPECIFICATIONS

Acceptance Criteria (Given-When-Then)

Turning vague business requests into testable engineering behavior.

// vague requirement: "Users should be able to reset password"
SCENARIO #1: HAPPY PATH

Valid Reset Request

GIVEN a registered user email
WHEN a valid reset request is submitted
THEN generate hashed token with 15m TTL & dispatch email
→ Influences API Handler & Redis TTL
SCENARIO #2: EXPIRED TOKEN

Expired Reset Attempt

GIVEN an expired or revoked reset token
WHEN the user submits new password
THEN reject request with 400 Bad Request (`TOKEN_EXPIRED`)
→ Influences Zod & DB Token validation
SCENARIO #3: BOT ATTACK

Rate Limit Exceeded

GIVEN repeated reset attempts within 60s
WHEN rate limit threshold is exceeded
THEN throttle further attempts with HTTP 429 Too Many Requests
→ Influences Redis Rate Limiter middleware
API ARCHITECTURE

API Protocols & Communication Styles

Knowing when to use REST, GraphQL, gRPC, WebSockets, SSE, or Webhooks.

// right protocol for the right boundary

REST

DECISION MATRIX
PROS:

Universal browser support, HTTP caching (ETags), simple status codes.

CONS & LIMITS:

Over-fetching / under-fetching entity graphs.

IDEAL WHEN:

Building clean resource-oriented endpoints with standard HTTP caching.

HTTP MECHANISMS

Advanced HTTP & Resiliency Patterns

// beyond basic GET / POST
IDEMPOTENCY & OPTIMISTIC LOCKING
// 1. Idempotent Deduplication (Redis key 24h TTL)
const idempotencyKey = req.headers["x-idempotency-key"];
const cachedResponse = await redis.get(`idemp:${idempotencyKey}`);
if (cachedResponse) return res.json(JSON.parse(cachedResponse));

// 2. Optimistic Concurrency Control (Version Check)
const updatedRows = await db("accounts")
  .where({ id: accountId, version: currentVersion })
  .update({ balance: newBalance, version: currentVersion + 1 });

if (updatedRows === 0) throw new ConcurrentUpdateConflictError();
ABORTSIGNAL TIMEOUTS & ETAG CACHING
// 1. Timeout external requests after 5000ms
const response = await fetch("https://gateway.payment.com", {
  signal: AbortSignal.timeout(5000)
});

// 2. ETags & Conditional 304 Not Modified
const etag = generateHash(responseData);
if (req.headers["if-none-match"] === etag) {
  return res.status(304).end(); // Zero payload transfer
}
res.setHeader("ETag", etag);
DOMAIN MODELING

Turn Nouns into Data. Turn Verbs into Behavior.

Extracting domain entities and relationships from real-world business requirements.

// requirement: "Customer places an order"
1. NOUN & VERB EXTRACTION
NOUNS (Data Entities):
CustomerOrderProductOrderItemPayment
VERBS (Business Behaviors):
places()calculates total()deducts stock()verifies payment()
2. ENTITY RELATIONSHIP & OWNERSHIP GRAPH
Customer (1)
   └── Order (1..N)
        ├── OrderItem (1..N)
        │      └── Product (1)
        └── Payment (1)

Domain modeling dictates table foreign keys, invariant rules, and cascade deletes.

SYSTEM ARCHITECTURE

Architecture Is a Decision, Not a Buzzword

// picking the right pattern for team scale & constraints

1. Layered Architecture

Strict separation of HTTP, business logic, and database access

// simple, understandable, perfect default for standard business applications
Controller → Service → Repository → Database
WHEN TO USE

Monolithic applications, REST APIs, domain-driven CRUD apps.

BENEFITS

Easy to reason about, low overhead, clear code organization.

TRADE-OFFS

Can encourage heavy database reliance if service logic spills into SQL.

ARCHITECTURE DECISION RECORDS

Why Did I Choose This Architecture? (ADRs)

Engineering means choosing trade-offs, not collecting buzzwords.

// documented rationale & trade-offs
DECISION RECORD

ADR-001: SQLite (WAL Mode) vs PostgreSQL for Local Portfolio Storage

DECISION MADE:

Use SQLite with Write-Ahead Logging (WAL) via `better-sqlite3`.

CONTEXT & NEED:

Portfolio site needs lightning-fast reads without heavy external database server overhead.

PRIMARY REASON:

Zero operational configuration, single-file deployment, sub-millisecond local reads, full SQL query support.

ACCEPTED TRADE-OFF:

Limited concurrent write throughput (not an issue for portfolio read workloads).

ENGINEERING RULES

Coding Principles (Rules I Follow)

// pragmatism > dogma
#1 Single Responsibility (SRP)

A module should have one clear reason to change.

Keep HTTP routing, business calculations, and SQL queries in separate files.

#2 Separation of Concerns

HTTP → Business Logic → Data Access.

Never write raw database SQL directly inside API request handlers.

#3 Dependency Inversion

High-level policy should not depend on low-level details.

Core business logic imports interfaces, not specific cloud SDK packages.

#4 DRY (Don't Repeat Yourself)

"DRY doesn't mean turning 2 lines of code into a 400-line generic framework."

Avoid duplication, but prefer duplicate code over the wrong abstraction.

#5 KISS (Keep It Simple, Stupid)

Prefer the simplest design that satisfies requirements.

Don't build complex Kubernetes cluster configs when a single VM works.

#6 YAGNI (You Aren't Gonna Need It)

"Future scale is not a feature request."

Don't write infrastructure for hypothetical features that don't exist yet.

#7 Composition Over Inheritance

Compose small reusable functions.

Chain focused middleware instead of inheriting deep class hierarchies.

#8 Explicit Over Clever

"If I need a decoder ring to review your function, something went wrong."

Readable, obvious code wins over hyper-clever one-liners every time.

QUERY PERFORMANCE

Why Is This Query Taking 4.2 Seconds? (EXPLAIN ANALYZE)

Senior backend engineers profile execution plans with EXPLAIN ANALYZE instead of adding more servers.

QUERY MODE:
EXECUTED SQL QUERY:SELECT * FROM orders WHERE customer_id = 'c_8810' AND status = 'PENDING' ORDER BY created_at DESC;
POSTGRES EXPLAIN ANALYZE OUTPUT:
Seq Scan on orders  (cost=0.00..18450.00 rows=420 width=128) (actual time=12.400..4185.320 ms)
  Filter: ((customer_id = 'c_8810'::uuid) AND ((status)::text = 'PENDING'::text))
  Rows Removed by Filter: 4,999,580
Sort  (cost=18490.12..18491.17 rows=420 width=128) (actual time=4198.100..4202.400 ms)
  Sort Key: created_at DESC
Execution Time: 4205.80 ms   <-- ⚠️ 4.2 SECONDS (Table Scan across 5,000,000 rows!)

Root Cause: Missing index forces PostgreSQL to read 5 Million disk blocks sequentially into RAM.

DATABASE INTERNALS

PostgreSQL & Relational DB Mechanics

// MVCC, PgBouncer, row locks & isolation levels
MVCC (Multi-Version Concurrency)

PostgreSQL writes new tuple versions on UPDATE instead of locking readers, keeping reads non-blocking.

ISOLATION LEVELS

Read Committed (default), Repeatable Read (phantom read protection), Serializable (strict serializability).

PGBOUNCER POOLING

Transaction-level connection pooling prevents backend process memory exhaustion under 10,000 clients.

NOSQL LANDSCAPE

NoSQL Placement Matrix

Document (MongoDB)Flexible schema, JSON catalogs
Key-Value (Redis)Sub-ms caching, rate limits, sessions
Wide-Column (Cassandra)High write throughput, timeseries
CACHING PATTERNS

Redis Caching & Stampede Defense

Cache-Aside (Lazy Loading)App checks Redis → DB on miss → Populates Redis
Stampede ProtectionDistributed Mutex Lock / Probabilistic Early Expiration
API DESIGN & CONTRACTS

An API is a Contract

Predictable REST resource endpoints & uniform error payloads.

// predictable JSON schemas for client stability
RESOURCE-ORIENTED ROUTES
POST /ordersCreate new order (201)
GET /orders/:idRetrieve order spec (200)
PATCH /orders/:idUpdate status / address
DELETE /orders/:idCancel pending order
PREDICTABLE ERROR FORMAT CONTRACT
{
  "code": "ORDER_NOT_FOUND",
  "message": "Order does not exist or has been deleted.",
  "requestId": "req_99182374",
  "timestamp": "2026-08-17T14:50:00Z"
}

Includes request correlation IDs for instant distributed log tracing without leaking stack traces.

DATA INTEGRITY

Database Engineering & Layered Rules

"If the rule matters, don't rely on one layer to remember it."
EXAMPLE REQUIREMENT: "An order should never have a negative total."
LAYER 1Business Rule
LAYER 2Zod Validation
LAYER 3SQL CHECK (total >= 0)
INDEXINGB-Tree indexes on FKs & query predicates
LOCKINGSELECT FOR UPDATE row-level locks
TRANSACTIONSACID guarantees across multi-row mutations
CONSTRAINTSForeign keys & unique indexes
TRANSACTION THINKING

What Happens When Step 3 Fails?

Backend engineering involves handling failure scenarios, not just happy paths.

TOGGLE SIMULATED GATEWAY:
STEP 11. Create OrderStatus: PENDING
STEP 22. Reserve StockStock -1 (Locked)
STEP 33. Charge Payment402 Declined!
STEP 4 (RECOVERY)4. State ResolutionROLLBACK & Release Stock 🔄
FAILURE HANDLING POLICY:

❌ Transaction automatically aborts. Inventory reservation is released, order state is marked `PAYMENT_FAILED`, and no partial orphan records remain in PostgreSQL.

HTTP SEMANTICS

Errors Are Part of the Design

// precise HTTP status codes > generic 500 errors
400Validation
401Unauthenticated
403Unauthorized
404Not Found
409Conflict
429Rate Limited
500Server Error
MESSAGING & EVENT-DRIVEN

Messaging Platforms & Event Streams

Decoupling synchronous operations into scalable background streams.

// Kafka, RabbitMQ, SQS & Dead-Letter Queues
APACHE KAFKAHigh-throughput log replay

Partitions, offsets, consumer groups, immutable log retention for event streaming.

RABBITMQComplex message routing

AMQP exchanges, topics, headers, acknowledgment, dead-letter exchanges (DLX).

AWS SQS / SNSCloud managed queues

Pub/Sub fanout (SNS) to isolated worker queues (SQS) with redrive policies.

DEAD-LETTER QUEUES (DLQ)Poison pill isolation

Failed messages isolated after 3 retries for developer inspection without blocking pipeline.

DISTRIBUTED CONSISTENCY

What Happens If Service A Succeeds but Service B Fails?

Handling distributed transactions across microservice boundaries without 2PC locking.

TRANSACTIONAL OUTBOX PATTERN

Guarantees that database state changes and message publishing succeed atomically in 1 transaction.

BEGIN TX → Update Order State → Insert Event to Outbox Table → COMMIT TX → Worker Relays Outbox to Kafka
CLOUD ARCHITECTURE

AWS Production Cloud Infrastructure Topology

Secure VPC network segmentation, edge CDN, load balancing, container orchestration, and KMS encryption.

// VPC private subnets & zero public DB exposure
Internet → CloudFront CDN → AWS WAF → Application Load Balancer (ALB) ↓ (Private Subnet) [ ECS / EKS Cluster ] (HPA) ├── ElastiCache (Redis Cluster) └── Aurora PostgreSQL (Multi-AZ Read Replicas)
ORCHESTRATION

Docker & Kubernetes Production Primitives

// multi-stage builds & health probes
MULTI-STAGE DOCKERFILEImage Optimization

Compiles TypeScript in build stage and copies only node_modules into dist, reducing image size from 1.2GB to 85MB.

PROBES & RESOURCE LIMITSLiveness & Readiness

Readiness probe `/api/health` ensures traffic is routed only after DB connections are warm.

HPA AUTO-SCALINGHorizontal Pod Autoscaler

Automatically scales Pod replicas from 3 to 30 when CPU utilization exceeds 70%.

DEVSECOPS PIPELINE

Automated Security Gates

Code SAST (Snyk) Dependency Scan Container Scan (Trivy) DAST (OWASP ZAP) Deploy 🔒
TELEMETRY STACK

Logs, Metrics & Distributed Traces

Logs:OpenSearch / ELK / Loki + Correlation IDs
Metrics:Prometheus + Grafana dashboards
Tracing:OpenTelemetry + Jaeger distributed context propagation
VERIFICATION STRATEGY

The Backend Testing Pyramid

Multi-tiered test coverage from isolated unit functions to 5,000 user JMeter load tests.

// requirement → implementation → test → evidence
E2E (User Journeys)
API & Integration Tests (DB / Redis / Gateways)
Unit Tests (Business Logic & Schema Validation Rules)
QUALITY GATES

When Is a Feature Actually Finished?

Not just "the API works on localhost".

"Merged != finished."
DEFINITION OF DONE PROGRESS:10 / 12 COMPLETED
Clean Code & Types
Zod Input Validation
RBAC Authorization
Unit & Integration Tests
Predictable Errors
Structured Logs
Telemetry Metrics
OpenAPI Spec / Docs
Safe DB Migration
Security Review
Load Consideration
CI/CD Deployment
TELEMETRY

The 4 Pillars of Observability

// know what is happening before users report outages
LOGS"WHAT HAPPENED?"

Structured JSON events with request IDs and timestamps.

METRICS"HOW MUCH / OFTEN?"

Counters, histograms, and p99 latency distributions.

TRACES"WHERE DID IT HAPPEN?"

Distributed span propagation across microservices.

AUDIT EVENTS"WHO PERFORMED IT?"

Immutable security record of data mutations.

AUTOMATION

CI/CD Pipeline Flow

Git Push Lint Type Check Unit Tests Integration Tests Security Scan Build Deploy Health Check
SECRETS & ENVS

Environment Isolation

Dev → Staging → Production
Secrets in Vault / Cloud KMS (Never in Git)
"`.env` is not a password manager."
DEFENSIVE SECURITY

Security by Design & Threat Modeling

// security is an architectural constraint, not a patch
EXAMPLE REQUIREMENT: "Only account owners can edit their profile."
RequirementAuthorization RuleRBAC Ownership CheckService ValidationSecurity Test
THREAT MODELING PIPELINE
1. ASSET
2. THREAT
3. RISK
4. CONTROL
5. TEST
PULL REQUEST AUDIT

What I Look For in a PR

Experienced backend questions asked during code review before approving merge requests.

// reviewing for edge cases & failure modes
"What happens when this request runs twice?"
"Can another user access this resource?"
"What happens when the database is down?"
"Why does this query need to run inside a loop?"
"What happens under 10,000 records?"
"Are sensitive fields masked in logs?"
"Is the transaction boundary tight enough?"
"Is there an index supporting this WHERE query?"
"Will this payload pass schema validation?"
DECISION MATRIX

There Is No Perfect Architecture

// engineering means choosing trade-offs, not collecting tech
SQLITE (WAL)+ Zero config, sub-ms local reads- Limited multi-node concurrent writes
POSTGRESQL+ ACID transactions, JSONB, concurrency- Requires connection pool infra
REDIS CACHING+ In-memory speed, sub-ms responses- Cache invalidation complexity
THE ENGINEERING SUMMARY

"Before I write code, I ask better questions."

“Good code starts before the first line of code.”

"The hardest part isn't writing the endpoint. It's deciding what the endpoint is actually allowed to do."

PROD SYSTEMS & APIS

selected work (backend-focused case studies)

/* click any project to view technical post-mortem */

Sangamam Backend (Core Banking Ecosystem)

2025

Cooperative banking platform backend, 45+ REST APIs, multi-branch RBAC, two-phase fund transfers, double-entry General Ledger engine.

Node.jsExpressSequelizePostgreSQLRedisAWS ECSS3DockerTypeScriptSwagger
⚡ Engineering Problem:

Concurrent intra-bank wallet fund transfers caused double-debit balance mismatches under simultaneous mobile app requests.

// active case study below

E-Commerce Platform

2025

Modern e-commerce backend platform featuring dynamic multi-variant catalog, inventory reservation locks, persistent cart, and checkout flow.

Next.js 16Node.jsPostgreSQLPrismaStripe APIZodRedis
⚡ Engineering Problem:

Cart items lost image thumbnail URLs during checkout transitions due to mismatched nested product data structures.

// click to inspect post-mortem

HRMS Lite (hrms-v1)

2025

Lightweight Human Resource Management System, employee attendance tracking, role-based access control, and payroll calculations.

FastifyTypeScriptPostgreSQLPrismaRedisJWT
⚡ Engineering Problem:

Biometric punch-in devices burst 10,000 requests in a 15-minute window every morning at 9:00 AM, causing database locks on the attendance table.

// click to inspect post-mortem

Garment Production & B2B/B2C Invoice Generator

2025

B2B bulk order management, B2C invoice generation engine, automated tax rules, and fabric stock tracking.

Node.jsExpressPostgreSQLRedisBullMQZodPDFKit
⚡ Engineering Problem:

B2B bulk invoices with 50+ line items and custom tax tiers were causing PDF rendering to timeout and block concurrent API requests.

// click to inspect post-mortem

BloodLink (Life-Saving Platform)

2024

Real-time blood donation platform connecting donors with seekers, donor 90-day cooling period enforcement, and geo-location search.

Next.jsNode.jsFastifySocket.ioMongoDBRedisGoogle Cloud VPS
⚡ Engineering Problem:

Seekers were receiving duplicate emergency notifications when multiple nearby donors clicked 'Accept' simultaneously.

// click to inspect post-mortem

AI Interview Coach (LLMs / Antigravity / Ollama)

2025

Generative AI technical interview simulator with architect-level feedback, local LLM backup, and real-time streaming analysis.

Next.js 14AntigravityOllamaGemini ProAWS LambdaTypeScriptPostgreSQL
⚡ Engineering Problem:

Cloud LLM API rate limits caused interview evaluation drops during peak user practice sessions.

// click to inspect post-mortem

Collaborative Whiteboard

2024

Real-time room-based whiteboard, Fastify WebSocket server, Redis room state persistence, and canvas sync.

FastifyNext.jsWebSocketsSocket.ioRedisPostgreSQLDockerPM2
⚡ Engineering Problem:

Broadcast lag increased linearly as room participant count grew above 50 concurrent drawers.

// click to inspect post-mortem

Add Secondary Devices with QR Auth

2024

Multi-device authentication system using QR codes, 1-hour session timeout, 3-minute QR validity, and Redis session caching.

Next.jsNode.jsPostgreSQLCryptoJSJWTRedisDockerPM2
⚡ Engineering Problem:

Replay attacks were possible if a generated QR code image was photographed and scanned after the primary user logged off.

// click to inspect post-mortem

Marriage Registration Module (Civil Court e-Services)

2024

Government civil court e-services module, multi-tier approval workflow (Assistant -> Sub-Registrar -> OSR eKYC), and 30-day objection notice automation.

Node.jsExpressPostgreSQLKnex.jsRedisCron
⚡ Engineering Problem:

Applications entering the 30-day objection period required precise automated status transitions without missing notice deadlines.

// click to inspect post-mortem

Premutation & MP Land Transaction Module (MPWebGIS)

2024

MP Government land transaction system, GIS-based land partition selection, real-time double-transaction checks, 48h payment validation cron.

Node.jsExpressPostgreSQLKnex.jsRedisCronGIS API
⚡ Engineering Problem:

Users selected land partitions but abandoned payment, locking land parcels indefinitely and preventing legitimate buyers from purchasing.

// click to inspect post-mortem
POST-MORTEM CASE STUDY

Sangamam Backend (Core Banking Ecosystem)

WHAT BROKE

Non-atomic database updates allowed two overlapping debit requests to evaluate balance checks simultaneously.

WHAT I CHANGED

Implemented a two-phase commit pattern in PostgreSQL transactions combined with Redis atomic rate-limiting and row-level locking.

WHY I CHOSE IT

Sequelize ORM transactions with PostgreSQL guaranteed ACID compliance, while AWS ECS provided zero-downtime auto-scaling.

WHAT I LEARNED

"Financial transactions require zero trust in timing; every balance mutation must be atomic and audit-logged."

CORE BACKEND RESPONSIBILITIES IN THIS PROJECT:
  • Architected and shipped customer & agent app backends with Aadhaar/MPIN auth and strict device binding
  • Engineered intra-bank fund transfer flow utilizing two-phase commit pattern and Redis rate-limiting
  • Built comprehensive 4,000+ line reporting engine and full double-entry General Ledger (GL) system from scratch
  • Orchestrated containerized microservices on AWS ECS with S3 compliance document storage
TOOLING PHILOSOPHY (29 / 29 SHOWING)

things I learned the hard way (the complete backend tool wall)

/* tools aren't badges; they are solutions to past disasters */

# Node.js & Event Loop

FRAMEWORK

"The workhorse runtime engine powering asynchronous, high-concurrency backend services."

Asynchronous non-blocking libuv event loop architecture
Single-threaded event-driven non-blocking I/O execution
High-throughput I/O bound REST and WebSocket APIs
💬 Single-threaded until you realize event-driven non-blocking I/O runs circles around synchronous multi-threading for I/O bound tasks.

# TypeScript

LANG & SPEC

"Static type checker ensuring runtime payload shape errors fail during build time instead of 3 AM production calls."

Strict type safety & interface compilation
Zod runtime schema type inference
Shared DTO contracts between backend and client
💬 Because 'undefined is not a function' isn't a surprise party you want at 3 AM in production.

# Fastify

FRAMEWORK

"Express's faster, schema-driven cousin engineered for raw API throughput and low latency."

Sub-millisecond HTTP routing overhead with Radix Tree matcher
Ajv JSON schema validation & fast serialization
Plugin encapsulation architecture preventing context leaks
💬 Why waste 5ms in routing middleware when Fastify does it in 0.2ms?

# Express.js

FRAMEWORK

"The classic, un-opinionated backend web framework for rapid REST API development."

Battle-tested REST route middleware pipeline
Flexible request/response transformation handlers
Universal NPM middleware ecosystem
💬 Old faithful: older than most JS frameworks, but still running half the internet.

# NestJS

FRAMEWORK

"Enterprise TypeScript framework enforcing clean architectural boundaries across large backend engineering teams."

Modular architecture with Dependency Injection
Decorators & TypeScript metadata reflection
Automated OpenAPI (Swagger) documentation generation
💬 For when your Node project grows so large that it starts craving Angular-style enterprise structure.

# PostgreSQL

DATABASE

"The rock-solid relational database of choice for financial ledgers, transactional ledgers, and complex queries."

ACID-compliant multi-version concurrency control (MVCC)
B-Tree, GIN, and Partial indexing strategies
Row-level locks (SELECT FOR UPDATE) and JSONB support
💬 Because your user's wallet balance shouldn't be an eventually-consistent guess.

# SQLite & Write-Ahead Logging

DATABASE

"Zero-network overhead embedded database that runs in-process with ultra-fast responses."

Zero-configuration embedded SQL database
Write-Ahead Logging (WAL) concurrent read performance
Sub-millisecond local disk & memory queries
💬 No network roundtrip means your database queries run faster than your frontend renders.

# MongoDB

DATABASE

"Document database ideal for rapidly changing document schemas and unstructured audit logs."

Document-oriented NoSQL storage
Flexible JSON-like schema structures
Aggregation pipeline framework for log data
💬 Schema-less is a dream until your code has to parse 4 different shapes of the same document.

# Redis & BullMQ

CACHE / QUEUE

"Ultra-fast in-memory cache and queue engine that sits in front of databases to handle traffic bursts."

In-memory key-value data structures with TTL eviction
Atomic distributed locks (SETNX) for race condition guards
Background worker queue handling with BullMQ
💬 Caching is easy until cache invalidation and distributed race conditions enter the chat.

# Apache Kafka

CACHE / QUEUE

"Event streaming backbone for publishing microservice events asynchronously with zero message loss."

Distributed commit log event streaming platform
Partitioned consumer groups for horizontal scaling
High-throughput asynchronous message pub/sub
💬 When HTTP webhooks just aren't durable enough for your millions of real-time event logs.

# Elasticsearch

DATABASE

"Dedicated search engine for instant full-text search across millions of complex records."

Distributed Lucene-based search engine
Inverted index architecture for full-text queries
High-volume log aggregation & analytics
💬 Because 'SELECT * FROM table WHERE text LIKE %query%' is a crime against database servers.

# AWS Cloud (ECS, S3, Lambda)

INFRA / DEVOPS

"Cloud infrastructure platform providing resilient compute, elastic scaling, and compliance storage."

AWS ECS container orchestration & auto-scaling groups
AWS S3 encrypted document vault & pre-signed URLs
AWS Lambda serverless event handlers
💬 The cloud is just someone else's server, but with auto-scaling and a monthly bill surprise.

# Docker & Containers

INFRA / DEVOPS

"Packages application dependencies into isolated containers so code runs identically anywhere."

Multi-stage container builds isolating application runtimes
Docker Compose local stack orchestration
Environment parameter standardization across stages
💬 'Works on my machine' -> Docker -> 'Now we ship your machine to production'.

# Kubernetes (K8s)

INFRA / DEVOPS

"Container orchestration system that keeps microservice clusters healthy and autoscaled."

Automated pod deployment & rolling update management
Horizontal Pod Autoscaling (HPA) based on CPU/Memory load
Self-healing container health checks and ingress routing
💬 100 YAML files later, your single container auto-scales like magic.

# Terraform

INFRA / DEVOPS

"Defines cloud servers, networks, and databases as version-controlled code rather than manual UI clicks."

Declarative Infrastructure as Code (IaC) configuration
State management & plan execution diffs
AWS VPC, RDS, and ECS infrastructure provisioning
💬 Why click buttons in AWS console when you can describe an entire datacenter in code?

# Ollama & Local LLMs

AI FRONTIER

"Runs AI language models locally on dedicated hardware for private, rate-limit-free AI workflows."

Local open-weight LLM runtime (Llama 3, DeepSeek, Qwen)
GPU node model hosting with GGUF quantization
Zero-cloud dependency offline fallback pipelines
💬 Because third-party cloud AI APIs will return 429 Too Many Requests right when your demo starts.

# Antigravity & Agentic Frameworks

AI FRONTIER

"Framework for designing multi-step AI agents that can reason, run commands, and execute code safely."

Autonomous agentic workflow orchestration & state loops
Structured prompt engineering & multi-step tool execution
Resilient error recovery & agentic decision trees
💬 Bridging the gap between human prompt intent and autonomous machine code execution.

# Apache JMeter

LOAD TEST

"Load testing tool used to hammer server APIs with high concurrency before actual users do."

Simulating 5,000+ concurrent user request waves
Measuring p50, p95, p99 latency distributions & error rates
Connection pool saturation & bottleneck discovery
💬 Because production users don't wait politely in line to hit your backend endpoints.

# Postman & Newman

API TOOLING

"The primary environment for crafting, testing, and documenting HTTP requests against raw backend routes."

REST & GraphQL API endpoint payload verification
Automated collection runner scripting via Newman
Environment variable injection & JWT auth testing
💬 The backend developer's true frontend interface.

# Burp Suite

SECURITY

"Security tool for intercepting and inspecting raw HTTP traffic to catch authorization bypasses."

Interception proxy for HTTP request/response tampering
Penetration testing payload manipulation
RBAC authorization bypass & security verification
💬 Proving that client-side validation is just a polite suggestion to an attacker.

# Zod Schema Validation

API TOOLING

"Validates incoming HTTP request bodies and parameters against strict schemas before executing business logic."

TypeScript-first static & runtime schema declaration
Strict input parsing & automatic error formatting
Inference of static TypeScript types from validation schemas
💬 Never trust user input; validate it at the gate before it breaks your DB query.

# WebSockets & Socket.io

INFRA / DEVOPS

"Real-time bi-directional messaging protocol for live push notifications and multiplayer collaboration."

Full-duplex real-time TCP socket connections
Room-based event broadcasting with sub-10ms latency
Heartbeat monitoring & automatic connection reconnection handling
💬 Polling every second is so 2010; WebSockets keep the connection open with sub-10ms events.

# Prisma & Sequelize & Knex

DATABASE

"Database abstraction tools providing type-safe querying and automated database schema migrations."

Type-safe ORM query generation & schema migrations
Relational mapping across complex foreign key structures
Raw SQL query builder flexibility with Knex.js
💬 Hiding raw SQL until you need to optimize a 5-way JOIN query with EXPLAIN ANALYZE.

# JWT & CryptoJS Security

SECURITY

"Cryptographic token system for handling stateless user sessions securely across microservices."

Stateless signed JSON Web Tokens for authorization
AES-256 payload encryption & HMAC signature verification
Redis token revocation blacklists for instant logout
💬 Stateless sessions are great until you need to revoke a compromised token immediately.

# PM2 & Linux Systemd

INFRA / DEVOPS

"Process management suite keeping backend server instances alive 24/7 on Linux VPS nodes."

Node.js process cluster mode for multi-core scaling
Automatic process restart on uncaught exceptions
Built-in log rotation and memory cap monitoring
💬 Keeping your Node process running even when an unhandled promise rejection tries to kill it.

# Python & Computer Vision

LANG & SPEC

"Versatile language used for backend automation scripts, computer vision processing, and AI integrations."

Scripting automation & data parsing pipelines
Pillow (PIL) pixel-level image processing & background removal
AI model embedding script integration
💬 When you need a 10-line script to manipulate 1,000 images or process machine learning embeddings.

# GraphQL

API TOOLING

"API query interface allowing clients to request exact fields, eliminating REST over-fetching."

Declarative field selection query language
Single HTTP endpoint consolidating nested data entities
Strongly-typed schema definition language (SDL)
💬 Solving REST over-fetching by giving frontend developers full query power over your DB schema.

# Git & GitHub Actions

INFRA / DEVOPS

"Version control and automated release engine ensuring every code push is tested and deployed safely."

Distributed version control & non-linear branching strategy
Automated CI/CD pipelines for linting, testing, and container builds
Pull request status checks & branch protection rules
💬 Because `git push --force` to main branch is not a valid continuous deployment strategy.

# Vitest & Jest

API TOOLING

"Automated test frameworks verifying backend functions, database queries, and API routes before shipping."

High-speed unit and integration test runner
Mocking HTTP routes, database pools, and external APIs
Code coverage reporting & snapshot testing
💬 Tests take 10 seconds to run; debugging un-tested production bugs takes 10 hours.
INTERACTIVE SIMULATORAPACHE JMETER LOAD SUITE

"What happens when 5,000 people hit the API at once?"

STATE:🙂WARMING
CONCURRENT USERS WAVE:🙂1,000 req/sec
😎 100 (Chilling)🙂 1,500 (Cruising)😅 3,500 (Sweating)😱 5,000 (MELTDOWN!)
AVG LATENCY32 ms
p99 LATENCY85 ms
CPU LOAD32%
ERRORS0%
CPU Core Stress:32%
DB Connection Pool (Max 100):28%
NODE_01
🙂
🙂"Thread pool active. Database indexes operating smoothly."
🛡️ DEFENSIVE ARCHITECTURE

things attackers notice before users do (security learning 🕵️‍♂️)

/* defensive security principles & interactive RBAC engine */

🔐 INTERACTIVE RBAC MATRIX EVALUATOR

// roles are not vibes 🙅‍♂️
1. SELECT USER ROLE (IDENTITY 👤):
2. SELECT ACTION INTENT (PERMISSION ⚡):
3. SELECT TARGET RESOURCE (DATABASE OBJECT 🗄️):
🛡️ BACKEND RBAC EVALUATORHTTP 200
🛡️😎
EVALUATION OUTCOME
ALLOWED ✅

Role 'EDITOR' explicitly grants 'UPDATE' permission on resource 'Project'.

📜 [SECURITY AUDIT] Timestamp=2026-08-17T14:50:00.000Z Role=EDITOR Action=UPDATE Resource=Project Outcome=ALLOWED_200
💉SQL INJECTION (SQLi)

Untrusted string concatenation transforms user input into executable SQL code inside the database engine.

❌ `SELECT * FROM users WHERE name = '` + input + `'`
✅ `SELECT * FROM users WHERE name = $1`
// parameterize everything 🔒
🔑AUTHENTICATION VS AUTHORIZATION

Authentication verifies identity. Authorization verifies permissions. Never mix the two.

User authenticated ✅ (JWT signature valid)
User authorized for `/admin/delete` ❌
// knowing who you are != touching everything 🚫
🚦RATE LIMITING & THROTTLING

Protect API infrastructure against credential stuffing, brute force, and runaway scraping bots.

Client → API Gateway → Redis Counter
Headers: X-RateLimit-Remaining
// rate limits save servers from bot waves 🤖
SYSTEM DESIGN & FLOWS

how the backend actually talks to itself (architecture playground)

/* click any step to trace data payload execution */

1. Request Lifecycle

From HTTP payload to SQL row query and JSON response

// step-by-step execution tracer
STEP #1Client Browser / Curl
STEP #2API Gateway
STEP #3Zod Schema Validation
STEP #4JWT Auth & RBAC Check
STEP #5Service & PostgreSQL
STEP #6HTTP 201 Created Response
STEP 1: CLIENT BROWSER / CURL

Sends HTTP POST /api/orders with JSON payload.

DATA PAYLOAD / STATE:{ "item": "Coffee", "qty": 2 }
TECH & RUNTIME

backend stack (human descriptions, no logo wall)

/* grouped by architectural purpose */

Languages

`TypeScript`

"strict types so runtime bugs hit the compiler, not production users."

`SQL`

"relational queries, indexing strategies, and transactional locks."

`JavaScript / ESNext`

"async/await, promises, and Node.js event loop mechanics."

Backend & APIs

`Node.js`

"non-blocking I/O runtime powering fast API gateways."

`Fastify`

"fast API framework without turning everything into middleware soup."

`Next.js 16 App Router`

"Server Components, Route Handlers, and backend-for-frontend APIs."

`REST APIs`

"clean OpenAPI specs, strict payload validation, and HTTP status codes."

Databases & Caching

`PostgreSQL`

"the gold standard relational DB with ACID guarantees and JSONB support."

`SQLite (WAL)`

"lightning-fast local embedded DB for zero-latency local state."

`Redis`

"in-memory caching, sliding window rate limits, and Pub/Sub queues."

Testing & Performance

`Apache JMeter`

"stressing APIs with 5,000 concurrent requests before launch."

`Postman`

"API payload exploration, header testing, and collection suites."

`Vitest / Jest`

"unit tests and integration route testing with mock databases."

Defensive Security

`Burp Suite`

"intercepting HTTP requests to test client-side security assumptions."

`RBAC & OWASP`

"role permission bitmasks, SQLi parameterized queries, and CORS."

`Rate Limiting`

"protecting endpoints from brute-force & denial-of-service bot waves."

Infrastructure & DevOps

`Docker`

"reproducible Linux containers so 'works on my machine' works everywhere."

`Linux / Bash`

"grep, tailing logs, systemd service units, and SSH remote administration."

`Git & CI/CD`

"clean commits, GitHub Actions automated build and test pipelines."

ARCHITECTURAL RATIONALE30 TECH DECISIONS

what I use vs why I use it (engineering decision matrix)

SHOWING 29 OF 29 ARCHITECTURAL TOOLS

Node.js & Event Loop

FRAMEWORK
🛠️ WHAT I USE (THE STACK):

High-concurrency event-driven server runtime

⚡ WHY I USE IT (THE PROBLEM SOLVED):

"How to process thousands of non-blocking I/O API connections on minimal memory footprint."

CATEGORY: FRAMEWORK✓ VERIFIED PROD TOOL

TypeScript

LANG & SPEC
🛠️ WHAT I USE (THE STACK):

Type safety & developer tooling overlay for JavaScript

⚡ WHY I USE IT (THE PROBLEM SOLVED):

"Catching contract mismatches and null pointer exceptions before code ever touches staging."

CATEGORY: LANGUAGES✓ VERIFIED PROD TOOL

Fastify

FRAMEWORK
🛠️ WHAT I USE (THE STACK):

Ultra-high performance HTTP web framework

⚡ WHY I USE IT (THE PROBLEM SOLVED):

"Eliminating HTTP framework routing latency spikes under high burst traffic."

CATEGORY: FRAMEWORK✓ VERIFIED PROD TOOL

Express.js

FRAMEWORK
🛠️ WHAT I USE (THE STACK):

Standard REST API server framework

⚡ WHY I USE IT (THE PROBLEM SOLVED):

"Rapidly bootstrapping HTTP API routing pipelines with zero boilerplate."

CATEGORY: FRAMEWORK✓ VERIFIED PROD TOOL

NestJS

FRAMEWORK
🛠️ WHAT I USE (THE STACK):

Structured enterprise microservice architecture

⚡ WHY I USE IT (THE PROBLEM SOLVED):

"Preventing large backend codebases from degrading into unmaintainable spaghetti."

CATEGORY: FRAMEWORK✓ VERIFIED PROD TOOL

PostgreSQL

DATABASE
🛠️ WHAT I USE (THE STACK):

Relational data persistence & transactional ACID integrity

⚡ WHY I USE IT (THE PROBLEM SOLVED):

"Storing complex financial & relational business data with zero corruption risk."

CATEGORY: DATABASE✓ VERIFIED PROD TOOL

SQLite & Write-Ahead Logging

DATABASE
🛠️ WHAT I USE (THE STACK):

Embedded lightweight SQL storage

⚡ WHY I USE IT (THE PROBLEM SOLVED):

"Providing zero-latency, file-based relational storage without running a standalone DB server."

CATEGORY: DATABASE✓ VERIFIED PROD TOOL

MongoDB

DATABASE
🛠️ WHAT I USE (THE STACK):

Flexible document storage

⚡ WHY I USE IT (THE PROBLEM SOLVED):

"Storing unstructured or dynamically evolving document attributes without running schema migrations."

CATEGORY: DATABASE✓ VERIFIED PROD TOOL

Redis & BullMQ

CACHE / QUEUE
🛠️ WHAT I USE (THE STACK):

In-memory high-speed data store & queue buffer

⚡ WHY I USE IT (THE PROBLEM SOLVED):

"Buffer high-frequency write traffic and answer repeated queries in under 0.5ms."

CATEGORY: CACHING✓ VERIFIED PROD TOOL

Apache Kafka

CACHE / QUEUE
🛠️ WHAT I USE (THE STACK):

Distributed event streaming log

⚡ WHY I USE IT (THE PROBLEM SOLVED):

"Decoupling microservices with durable, replayable event queues at massive throughput."

CATEGORY: CACHING✓ VERIFIED PROD TOOL

Elasticsearch

DATABASE
🛠️ WHAT I USE (THE STACK):

Full-text search & log analytics engine

⚡ WHY I USE IT (THE PROBLEM SOLVED):

"Executing sub-second searches across millions of un-structured document records."

CATEGORY: DATABASE✓ VERIFIED PROD TOOL

AWS Cloud (ECS, S3, Lambda)

INFRA / DEVOPS
🛠️ WHAT I USE (THE STACK):

Resilient cloud compute & object storage

⚡ WHY I USE IT (THE PROBLEM SOLVED):

"Deploying microservices with automated failover, auto-scaling, and secure storage."

CATEGORY: INFRASTRUCTURE✓ VERIFIED PROD TOOL

Docker & Containers

INFRA / DEVOPS
🛠️ WHAT I USE (THE STACK):

Runtime environment containerization

⚡ WHY I USE IT (THE PROBLEM SOLVED):

"Eliminating environment drift bugs between development laptops and production servers."

CATEGORY: INFRASTRUCTURE✓ VERIFIED PROD TOOL

Kubernetes (K8s)

INFRA / DEVOPS
🛠️ WHAT I USE (THE STACK):

Production container cluster orchestration

⚡ WHY I USE IT (THE PROBLEM SOLVED):

"Automating zero-downtime rolling updates, pod restarts, and load balancing across multi-node clusters."

CATEGORY: INFRASTRUCTURE✓ VERIFIED PROD TOOL

Terraform

INFRA / DEVOPS
🛠️ WHAT I USE (THE STACK):

Infrastructure as Code (IaC) automation

⚡ WHY I USE IT (THE PROBLEM SOLVED):

"Making infrastructure provisioning repeatable, audit-logged, and peer-reviewable."

CATEGORY: INFRASTRUCTURE✓ VERIFIED PROD TOOL

Ollama & Local LLMs

AI FRONTIER
🛠️ WHAT I USE (THE STACK):

Local private LLM inference engine

⚡ WHY I USE IT (THE PROBLEM SOLVED):

"Integrating generative AI into backend applications without cloud rate limits or privacy leaks."

CATEGORY: AI_FRONTIER✓ VERIFIED PROD TOOL

Antigravity & Agentic Frameworks

AI FRONTIER
🛠️ WHAT I USE (THE STACK):

Agentic AI orchestration & tool-calling framework

⚡ WHY I USE IT (THE PROBLEM SOLVED):

"Executing complex multi-step reasoning tasks without human intervention loops."

CATEGORY: AI_FRONTIER✓ VERIFIED PROD TOOL

Apache JMeter

LOAD_TESTING
🛠️ WHAT I USE (THE STACK):

API load & stress testing engine

⚡ WHY I USE IT (THE PROBLEM SOLVED):

"Uncovering database deadlocks and memory leaks under simulated extreme traffic."

CATEGORY: LOAD_TESTING✓ VERIFIED PROD TOOL

Postman & Newman

API_TESTING
🛠️ WHAT I USE (THE STACK):

API verification & testing suite

⚡ WHY I USE IT (THE PROBLEM SOLVED):

"Verifying backend API responses and error codes independently of UI implementations."

CATEGORY: API_TESTING✓ VERIFIED PROD TOOL

Burp Suite

SECURITY
🛠️ WHAT I USE (THE STACK):

Web security & penetration testing proxy

⚡ WHY I USE IT (THE PROBLEM SOLVED):

"Identifying RBAC flaws, unparameterized queries, and header vulnerabilities before attackers do."

CATEGORY: SECURITY✓ VERIFIED PROD TOOL

Zod Schema Validation

API_TESTING
🛠️ WHAT I USE (THE STACK):

Runtime data contract & schema validation

⚡ WHY I USE IT (THE PROBLEM SOLVED):

"Preventing malformed request payloads from causing silent runtime bugs deep inside business logic."

CATEGORY: API_TESTING✓ VERIFIED PROD TOOL

WebSockets & Socket.io

INFRA / DEVOPS
🛠️ WHAT I USE (THE STACK):

Real-time bi-directional web protocol

⚡ WHY I USE IT (THE PROBLEM SOLVED):

"Pushing server events to connected clients instantaneously without polling overhead."

CATEGORY: INFRASTRUCTURE✓ VERIFIED PROD TOOL

Prisma & Sequelize & Knex

DATABASE
🛠️ WHAT I USE (THE STACK):

Type-safe database abstraction & ORM layers

⚡ WHY I USE IT (THE PROBLEM SOLVED):

"Bridging TypeScript code models with SQL databases safely and cleanly."

CATEGORY: DATABASE✓ VERIFIED PROD TOOL

JWT & CryptoJS Security

SECURITY
🛠️ WHAT I USE (THE STACK):

Stateless authentication & cryptographic security

⚡ WHY I USE IT (THE PROBLEM SOLVED):

"Authenticating user requests across distributed services without database session lookups on every request."

CATEGORY: SECURITY✓ VERIFIED PROD TOOL

PM2 & Linux Systemd

INFRA / DEVOPS
🛠️ WHAT I USE (THE STACK):

Production process management

⚡ WHY I USE IT (THE PROBLEM SOLVED):

"Ensuring zero downtime, automatic crash restarts, and full CPU core utilization."

CATEGORY: INFRASTRUCTURE✓ VERIFIED PROD TOOL

Python & Computer Vision

LANG & SPEC
🛠️ WHAT I USE (THE STACK):

Scripting, computer vision, and AI processing

⚡ WHY I USE IT (THE PROBLEM SOLVED):

"Handling complex image processing, computer vision algorithms, and AI scripting with ease."

CATEGORY: LANGUAGES✓ VERIFIED PROD TOOL

GraphQL

API_TESTING
🛠️ WHAT I USE (THE STACK):

Declarative API query layer

⚡ WHY I USE IT (THE PROBLEM SOLVED):

"Fetching complex nested relational data trees in a single client roundtrip without payload bloat."

CATEGORY: API_TESTING✓ VERIFIED PROD TOOL

Git & GitHub Actions

INFRA / DEVOPS
🛠️ WHAT I USE (THE STACK):

Version control & continuous integration

⚡ WHY I USE IT (THE PROBLEM SOLVED):

"Preventing code conflicts and automating test execution before code reaches production."

CATEGORY: INFRASTRUCTURE✓ VERIFIED PROD TOOL

Vitest & Jest

API_TESTING
🛠️ WHAT I USE (THE STACK):

Automated test execution suite

⚡ WHY I USE IT (THE PROBLEM SOLVED):

"Guaranteeing legacy code doesn't break when new features or refactors are merged."

CATEGORY: API_TESTING✓ VERIFIED PROD TOOL
CONTINUOUS TECHNICAL GROWTH

things I'm currently learning (learning roadmap)

MASTERY INDEX100% Complete
4 Mastered0 Building0 Exploring
1
TOPIC #1

Backend Fundamentals & HTTP Specification

MASTERED
WHAT I UNDERSTAND:

HTTP status codes, headers, method semantics (GET vs POST vs PUT vs PATCH vs DELETE), idempotent vs non-idempotent operations, body stream handling.

WHAT I STILL NEED TO EXPLORE:

HTTP/3 QUIC protocol details & custom HTTP proxying layer tuning.

ENGINEER NOTE:"HTTP specification reading changed my perspective on API contract design."
STATUS: MASTERED
2
TOPIC #2

API Architecture & Gateway Design

MASTERED
WHAT I UNDERSTAND:

REST principles, OpenAPI specs, validation schemas (Zod/TypeBox), routing overhead, middleware chains, error handling standardization.

WHAT I STILL NEED TO EXPLORE:

gRPC proto contracts and Protobuf serialization speed comparisons against JSON.

ENGINEER NOTE:"Clean route schemas eliminate 90% of invalid runtime payload bugs."
STATUS: MASTERED
3
TOPIC #3

Database Design & SQL Performance

MASTERED
WHAT I UNDERSTAND:

Relational schema design, 3NF normalization, foreign key constraints, B-Tree index mechanics, EXPLAIN query planner output, connection pooling.

WHAT I STILL NEED TO EXPLORE:

Sharding algorithms and PostgreSQL multi-region active-active logical replication.

ENGINEER NOTE:"A missing index on a 2-million row table is the fastest way to bring down an API server."
STATUS: MASTERED
4
TOPIC #4

Caching Patterns & Memory Stores

MASTERED
WHAT I UNDERSTAND:

Cache-aside strategy, write-through caching, TTL policy selection, cache stampede prevention, Redis memory data types.

WHAT I STILL NEED TO EXPLORE:

Memcached vs Redis cluster key distribution hashing under node failures.

ENGINEER NOTE:"Cache invalidation is a business logic problem, not just a key deletion call."
STATUS: MASTERED
RAW BATTLE OBSERVATIONS

notes from the backend trenches (developer sticky wall)

/* developer observations collected over years of production outages */

security
RULE #1

"Never trust frontend validation."

Client-side validation is for UX. Backend validation is for security and data integrity. Anyone can send raw HTTP requests with curl or Postman.

VERIFIED IN PRODUCTION✓ Production Tested
database
RULE #2

"Indexes exist because databases also get tired."

Scanning 1,000,000 unindexed rows for every user search query turns your database server into a space heater.

VERIFIED IN PRODUCTION✓ Production Tested
security
RULE #3

"Authentication without authorization is just knowing someone's name."

Verifying WHO the user is doesn't mean they are allowed to read, edit, or delete the resource they requested.

VERIFIED IN PRODUCTION✓ Production Tested
architecture
RULE #4

"Burst traffic shouldn't hit relational databases directly."

Buffer high-frequency write traffic in Redis memory queues first, then flush in transactional batches to PostgreSQL.

VERIFIED IN PRODUCTION✓ Production Tested
SECURE DISPATCH GATEWAY

got a backend problem?

Tell me what is broken. I promise not to immediately blame DNS.

RATE LIMIT: 3 REQ/MIN