INDEX

シェク・ジュナエッド

Junaid Shaikh

Backend-focused Fullstack Developer

LOCATION
Mumbai, IN

I build backends that hold up under load: async job pipelines, type-safe APIs, and LLM workflows on a TypeScript / Node / Postgres stack. I care about the unglamorous parts, like idempotency, crash recovery, and data that stays correct when things go sideways.

Architected a production-bound AI platform backend from scratch. Now looking for the next hard problem.

00About

I'm a recent Information Technology grad from the University of Mumbai (8.38 CGPI) who fell hard for the backend. Most people see the UI; I think about what's happening behind it: the queue draining jobs, the transaction holding a write together, the worker recovering from a crash without losing state.

At Slaab.ai I architected the backend of an AI SEO platform from scratch and learned what "production-bound" actually demands. The rest of the time I'm on Linux breaking and fixing my own setup, chipping away at Japanese (JLPT N3), and grinding LeetCode while pretending it's fun.

Interests

Distributed SystemsAsync Job ProcessingAPI DesignDatabases & IndexingAI / LLM WorkflowsLinux & Self-hosting
Based in
Mumbai, India
Focus
Scalable Backend systems
Daily driver
Linux (Fedora)
Learning
日本語 (JLPT N3)
Education
BSc IT, University of Mumbai · 8.38 CGPI

01Experience

Where I've shipped real systems.

Slaab.ai

Full-Stack Developer
May 2025 – Jul 2025 · Remote, India

Architected and built the backend from scratch for a production-oriented AI SEO platform, designing domain-driven services, asynchronous execution pipelines, resilient integrations, and end-to-end type-safe workflows while delivering the complete frontend alongside the platform.

  • 01Architected the backend from scratch, defining a domain-driven TypeScript architecture spanning REST APIs, tRPC services, background workers, and shared validation contracts.
  • 02Designed a ten-stage agentic SEO audit workflow orchestrating LLM reasoning with Lighthouse, DataForSEO, Firecrawl, and security analysis to transform raw crawl data into prioritized, actionable recommendations.
  • 03Engineered a queue-backed asynchronous execution platform using BullMQ and Redis, decoupling long-running AI workloads from the request path with graceful shutdown, worker lifecycle management, PostgreSQL-backed crash recovery, and resumable workflows.
  • 04Built resilient integrations with external SEO services using tiered scraping (Firecrawl → Cheerio → Puppeteer), URL normalization, structured validation, retry strategies, and per-job API/token accounting for cost-aware execution.
  • 05Developed an AI content generation pipeline combining outline generation, structured writing, iterative humanization, persistent conversational memory, and tool-calling workflows powered by Mastra.ai.
  • 06Delivered the complete Next.js frontend with Supabase SSR authentication, TanStack Query, Zustand, resumable job polling, and active-job recovery, allowing users to seamlessly reconnect to long-running background processes.
  • 07Established compile-time API guarantees across frontend and backend using shared Zod contracts, eliminating client/server contract drift while maintaining strongly typed interfaces.
TypeScriptBunHonoNext.jsBullMQRedisPostgreSQL (Supabase)Mastra.aiDocker

03Skills

What I reach for when I build.

Languages
TypeScript / JavaScriptGoCore JavaC / C++ (basic)
Backend
Node.js / BunHono.jsExpress.jstRPCBullMQ / Redis QueuesMastra.ai (agents)WebSocket / Socket.io
Frontend
React.jsNext.jsTanStack QueryZustandRedux / RTKTailwind CSS
Data & Caching
PostgreSQLPrismaDrizzlesqlcMongoDBRedisSQLite
DevOps & Tooling
Docker / PodmanNginxAWS (EC2, ECS, S3, RDS, Lambda)Git / GitHubLinux / Bash

02Selected Work

Side projects where I got to make the hard calls myself: pagination strategy, real-time consistency, deployment, all of it.

001

Aug 2023 – Mar 2024

Discord-like Chat Platform

Distributed real-time communication platform

Architected a distributed real-time communication platform supporting servers, channels, direct messaging, media sharing, presence, voice/video calls, and fine-grained role-based authorization. Built as a production-oriented Turborepo monorepo with end-to-end type safety, horizontally scalable messaging infrastructure, and direct-to-object-storage media workflows.

  • 01Designed a Redis-backed Socket.IO cluster using pub/sub fan-out, heartbeat-driven presence tracking, and TTL-based offline detection, enabling horizontally scalable real-time messaging with eventually consistent presence across distributed application nodes.
  • 02Engineered optimistic messaging through TanStack Query cache synchronization and WebSocket acknowledgements, delivering instant create/edit/delete operations while eliminating polling and redundant API synchronization.
  • 03Implemented transactional messaging workflows with cursor-based pagination, deterministic ordering, and authorization enforced at the query layer to guarantee consistency under concurrent user activity.
  • 04Built secure direct-to-object-storage uploads using presigned S3 URLs, server-side media validation, upload policies, and OAuth asset re-hosting, completely removing backend file proxying from the request path.
  • 05Established compile-time API guarantees across frontend and backend through shared Zod contracts inside a Turborepo monorepo, preventing runtime contract drift and maintaining strongly typed interfaces.
  • 06Implemented a hybrid notification pipeline combining Socket.IO for online delivery and Web Push notifications for offline users, using service workers, VAPID authentication, subscription management, and delivery fan-out to ensure users never missed mentions or direct messages.
  • 07Containerized the platform using Docker, Nginx reverse proxy, and Grafana Loki centralized logging, providing production-ready deployment, observability, and stateless application services.
Next.jsTypeScriptSocket.ioHonoBunDrizzlePostgreSQLRedisLiveKitTanStack QueryZustandMinIO (S3)Docker
002

Jun 2026

Auditd

Tamper-Evident Multi-Tenant Audit Logging Service

Designed a tamper-evident, multi-tenant audit logging service that provides cryptographically verifiable append-only event storage for compliance workloads, emphasizing deterministic ordering, immutable persistence, and constant-memory verification.

  • 01Designed a cryptographically verifiable append-only audit log using chained HMAC-SHA256 hashes across audit records, making unauthorized modification immediately detectable and fully traceable.
  • 02Engineered an asynchronous ingestion pipeline using buffered channels, worker pools, exponential backoff retries, bounded queues, and graceful shutdown to guarantee reliable event persistence during service termination.
  • 03Enforced immutable storage through PostgreSQL RULEs while coordinating concurrent writes with tenant-scoped advisory locks, preserving deterministic hash ordering without introducing global contention.
  • 04Optimized audit retrieval using composite indexes, tenant-aware caching, connection pooling, and multi-dimensional filtering across actors, resources, actions, timestamps, and event types.
  • 05Implemented constant-memory streaming exports (CSV/JSONL) and complete audit-chain verification APIs capable of validating historical integrity without loading the entire dataset into memory.
GoGinPostgreSQLpgxsqlc
003

Nov 2023 – Dec 2023

Notion Lite

Collaborative Notion-style workspace

Built a collaborative Notion-style workspace supporting conflict-free real-time editing, live cursors, user presence, recursive document hierarchies, and secure publishing workflows using CRDT synchronization over Cloudflare's edge infrastructure.

  • 01Built a collaborative Notion-style workspace supporting real-time CRDT synchronization, live cursors, and user presence using Yjs with Cloudflare Durable Objects for distributed session coordination.
  • 02Designed infinitely nested document hierarchies using sixteen Convex serverless functions and composite indexes, enabling recursive archive/restore workflows, orphan-safe re-parenting, and ownership enforcement.
  • 03Implemented authenticated editing, secure publishing workflows, public preview routes, and server-side authorization, eliminating unauthorized document modification while maintaining seamless collaboration.
  • 04Delivered a feature-rich editor across sixty-plus React components including BlockNote editing, emoji/icon support, cover images, search, responsive layouts, and theme switching while preserving real-time consistency.
Next.jsTypeScriptConvexCloudflare WorkersDurable ObjectsClerkYjsBlockNote
004

Apr 2023 – Jun 2023

Game Explorer

Game-discovery SPA over the RAWG API

A responsive discovery app over RAWG's 880K+ game catalog, with multi-criteria filtering, infinite scroll, and a caching strategy that all but eliminates redundant fetches.

  • 01Multi-criteria filtering (genre, platform, sort, search) + infinite-scroll pagination across 25+ components and 11 custom hooks.
  • 02Pre-seeded genre/platform lists as TanStack Query initialData with a 24-hour staleTime, so filter sidebars get a zero-network first paint and never refetch within the cache window.
  • 03Refactored a prop-drilled tree to a Zustand store + a generic ApiClient<T> over Axios; route-level code-splitting, dynamic SEO, and a multi-stage Docker build (Bun → Nginx).
ReactTypeScriptZustandTanStack QueryAxiosRAWG APIDocker

Go REST API

Learning Go, the systems way

A compact REST API I built to get fluent in Go: clean handler structure, SQL over an ORM, containerized for deployment.

  • 01Idiomatic Go service with Gin, hand-written SQL against LibSQL, and a Dockerfile tuned for small images. A deliberate exercise in writing Go that reads like Go.
GoGinLibSQLDocker

Redis Rate Limiter

Published npm library · token bucket in Lua

A published npm package that does rate limiting the way it should be done: the entire token-bucket decision runs inside Redis as an atomic Lua script, so every check is a single network round-trip with no race conditions. Drop-in RateLimiter class plus optional Hono middleware.

  • 01Token bucket implemented end-to-end in a Redis Lua script: read state, refill by elapsed time, cap, and decrement all happen atomically server-side, which kills the check-then-act race that breaks naive limiters under concurrency.
  • 02One EVALSHA call per request: the script is cached in Redis, so each rate-limit check costs a single round-trip regardless of load, designed to sit on the hot path without becoming the bottleneck.
  • 03Framework-agnostic RateLimiter core with an optional Hono middleware that returns a clean 429 JSON response; fully typed, configurable maxRequests / windowSizeInSeconds, shipped as ESM via tsup.
  • 04Published to npm as @zjunaidz/rate-limiter, so it's installable, documented, and reusable in real projects, not a throwaway demo.
TypeScriptRedisLuaioredisHono.jstsupnpm

There's more on my GitHub, and I'm always mid-build on something bigger.