_{a change in one file, and everything it touches — the kind of thing my tooling reasons about}

Software Engineer building AI-powered developer tools and the infrastructure behind them. I work on RAG pipelines, knowledge graphs, and MCP servers — and on getting all of it running reliably inside other people's infrastructure.

• 🚀 Currently Engineering Manager @ Bytebell, building an AI platform for querying codebases, docs, and organizational knowledge
• 🏗️ Most of my time goes to distribution & deployment — shipping the same product as multi-tenant cloud SaaS and as a self-hosted enterprise package on Kubernetes, Docker Compose, and client-owned AWS / GCP
• 🔔 Built the email notification system (queued delivery + automatic retries) that keeps users updated on ingestion and deployment events without dropping a single alert
• 🧠 Deep in TypeScript, Node.js, Fastify, MongoDB, Neo4j, Pinecone, Redis, Docker, Kubernetes, AWS, LLMs & MCP
• 📫 Reach me at nit8339@gmail.com

• ☁️ Deployed the platform to client-owned servers on AWS and GCP, handling infrastructure setup and production issues end to end
• 📦 Built the containerized distribution pipeline — versioned release packaging, Docker Compose & Kubernetes deployment, licensing manifests in etcd, HAProxy routing, and signed image publishing to a private registry (AWS ECR) — so customers run the full stack in their own environment
• 🔄 Shipped a self-updating system service that pulls signed images from the private registry, keeping single-tenant on-premise installations current with no manual upgrade work from the customer
• 🔔 Built an email notification system for key events (ingestion completion, deployment / system status) with queued delivery and automatic retries, so notifications are retried instead of lost
• ⚡ Cut end-to-end processing time by 30% and infrastructure cost by 40% through pipeline batching and query optimization
• 🔐 Established multi-organization access control with RBAC, per-org data isolation, and super-admin roles — one deployment serving multiple enterprise teams
• 🔗 Built OAuth 2.0 integrations with GitHub, GitLab, and Google (GitHub Apps + GitLab REST APIs), letting organizations securely connect their accounts and repositories
• 🛠️ Resolved 150+ production issues across client deployments

• 🏛️ Built and shipped a hybrid B2B SaaS AI developer-knowledge platform — RAG and knowledge-graph search over code repositories and documents, delivered both as multi-tenant cloud and as a self-hosted enterprise package
• 📥 Developed the content ingestion system letting companies add Git repositories, websites, PDFs, and images, then query them in natural language through a Pinecone-backed RAG pipeline
• 🔌 Built the MCP server exposing retrieval to coding agents, with session management, persistent connections, and per-user API keys
• 🖥️ Shipped client surfaces including a VS Code extension and an Electron desktop app for managing locally running models
• 💬 Designed the chat system with per-user persistent context, letting users save and reuse their own knowledge across conversations

• 📚 Resolved 100+ student questions on competitive programming, data structures, and algorithms
• 🤝 Mentored 50+ students, maintaining a 4.8/5 average rating

Languages

AI & LLM

Backend

Databases

Infra & DevOps

Frontend

Auth & Integrations

Tools

Shipping software into environments you don't control is a different problem from shipping to your own cloud — no SSH when something breaks, every customer setup is different, and you still have to deliver updates, enforce licensing, and keep people informed when jobs finish or fail. Building and running that pipeline — self-hosted on Kubernetes and Docker Compose, with signed images, self-updating services, and a notification system that retries instead of dropping alerts — is the work I do best.

The other thing I obsess over is retrieval quality: closing the gap between a RAG demo that looks impressive and one that actually returns the right context on a large, messy codebase.