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 and deployment — shipping the same product as multi-tenant cloud SaaS and as a self-hosted enterprise package
- 🧠 Working with TypeScript, Node.js, Fastify, MongoDB, Neo4j, Pinecone, Docker, AWS, LLMs & MCP
- 🌱 Exploring AI agents, retrieval quality, and distributed systems
- 📫 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 stacks, deployment and licensing manifests in etcd, HAProxy routing, and signed image publishing to a private registry
- 🔄 Shipped a self-updating system service that keeps single-tenant on-premise installations current with no manual upgrade work from the customer
- ⚡ Cut end-to-end processing time by 30% and infrastructure cost by 40% through pipeline batching and query optimization
- 🔐 Built multi-organization access control with RBAC, per-org data isolation, and super-admin roles
- 🛠️ 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
- 📥 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
- 📚 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. You can't SSH in when something breaks, every customer setup is different, and you still have to deliver updates and enforce licensing. Building that pipeline taught me more than anything else I've worked on.
I'm also interested in retrieval quality — the gap between a RAG demo that looks impressive and one that actually returns the right context on a large, messy codebase.
