Open to remote roles

Backend & AI Engineer · Ghaziabad, India

Mohit Kumar

I build production backend & AI systems — RAG pipelines, distributed infra, and the correctness underneath.

$ npm i token-bucket-limiter 

Mohit Kumar
01Selected work

Things I built and shipped

RAG AI Code Reviewer

Codebase-aware pull-request reviews via a from-scratch RAG pipeline

A GitHub App that reviews PRs against the repo's own patterns — it retrieves the top-5 most semantically similar functions so suggestions match established conventions instead of giving generic feedback.

  • Full RAG pipeline from scratch — function-level AST chunking with Tree-sitter, 1536-dim embeddings in pgvector, cosine-similarity retrieval. No third-party RAG framework.
  • Two-process architecture (webhook receiver + async worker) bridged by a pg-boss queue with idempotent, exactly-once processing under retries and network failure.
TypeScriptNode.jsPostgreSQLpgvectorTree-sitterOpenAIpg-boss

token-bucket-limiter

A distributed, Redis-backed rate limiter for Express — published on npm

An open-source token-bucket rate limiter that makes per-request decisions atomically inside Redis via a Lua script, so it stays correct across multiple Node processes and machines.

  • Atomic per-request decisions run inside Redis (Lua) — no race conditions under concurrency.
  • Published, versioned, and documented with a write-up on the design trade-offs.
TypeScriptRedisLuaExpressnpm
02About

Systems-first, correctness-obsessed

I'm a software engineer at Lunacal, a scheduling SaaS with 10,000+ users, where I own features end-to-end across a Turborepo / Next.js / tRPC / PostgreSQL stack.

I care most about correctness under concurrency — idempotent webhooks, atomic rate limiting, and consistent state across linked records. Outside work I build and publish my own systems, like a from-scratch RAG code-review pipeline and a Redis-backed distributed rate limiter live on npm.

Experience

  1. Software Engineer · Lunacal

    Sep 2025 — Present
    • Own features end-to-end for a scheduling SaaS with 10,000+ users across Turborepo, Next.js, tRPC and PostgreSQL.
    • Built recurring-events & multi-slot scheduling — sharing one meeting link across a recurring series and multi-session packages (up to 20 slots), engineering around Google Meet's per-link limit while keeping every linked booking consistent.
    • Shipped the end-to-end client portal, a configurable transactional-email system (AWS SES), a bidirectional HubSpot integration, GTM lifecycle tracking, and the Insights dashboard.
    • Published 3 no-code plugins — a live Framer Marketplace plugin plus WordPress and Webflow.
  2. Full Stack Developer · Infoware India

    Aug 2024 — Sep 2025
    • Built core modules including Inspections, Schedules, Tasks, and Assets for an enterprise inspection management platform using Next.js and TypeScript.
    • Collaborated with the client and engineering team to deliver production-ready functionality for a SafetyCulture-like platform.
    • Integrated REST APIs and implemented reusable UI components for complex business workflows.
  3. Frontend Developer Intern · ProjectBlaze

    Jan 2024 — Mar 2024
    • Built and styled responsive user-facing UI, translating designs into clean, reusable frontend components.
03Stack

Tools I reach for

Languages

TypeScriptJavaScriptPythonSQL

AI / LLM

RAG pipelinesEmbeddings & vector searchTree-sitter AST chunkingPrompt engineeringLLM orchestrationEval pipelinesMCP

AI Agents

LangChainLangGraphAI agentsMulti-agent systemsTool / function callingAgent memory

Frontend

ReactNext.jsTailwind CSSshadcn / RadixFramer MotionReact QueryZustandRedux

Backend

Node.jsExpress.jstRPCFastAPIPydanticZodREST APIs

Auth & Payments

OAuth / GoogleJWT & sessionsNextAuth (Auth.js)RazorpaySubscription billing

Systems

Distributed systemsRate limiting (Redis + Lua)Webhook idempotencyJob queues (pg-boss)Multi-tenancyAES-256-GCM

Databases

PostgreSQLpgvectorRedisMongoDB

DevOps & Tooling

DockerAWS (EC2, SES)AzureGitHub ActionsSentryTurborepoVercelGitLinux

Currently exploring

LLM fine-tuningMultimodal AI (text · image · audio)

Contact

Let's build something that ships.

Open to remote roles. Grab a time on my calendar, or drop me an email — I reply quickly.