About

I'm Shekhar, a senior software engineer at Pure Billion Technologies. I've owned backends end to end, including a wedding marketplace with real-time chat at 10k+ messages a day, Stripe Connect payment flows, and an automated multi-tenant migration system.

More and more of my work is the infrastructure under AI products: model routing, tool calling, retrieval, GPU inference and agent evaluation.

I also build NeatCode, an open-source engine that indexes TypeScript codebases into a SQLite code graph and serves it to coding agents over MCP.

At a glance

Name: Shekhar Kumar
Also known as: HITMANdark07 on GitHub
Role: Senior Software Engineer at Pure Billion Technologies
Focus: Backend & AI infrastructure: distributed systems, LLM platforms, GPU inference, agent evaluation
Based in: Bengaluru, India · works remotely
Open source: NeatCode (@neatcodeai/cli on npm) and mcp-evals
Experience: 4+ years shipping production systems in TypeScript, Go, Python and PostgreSQL
Contact: shekharkumar235838@gmail.com

Currently

as of October 2026

Building mcp-evals — an open-source eval harness for MCP servers. Its first real benchmark, NeatCode vs. the filesystem server across three models, is published.
Shipping NeatCode — v0.1.4 is on npm as @neatcodeai/cli, with installer and version fixes.
Senior Software Engineer at Pure Billion Technologies — working on AI infrastructure, agent evaluation and backend systems.

Experience

Aug 2022 – present
Senior Software Engineer, Pure Billion Technologies
Mumbai, IN · Remote
- Built NeatCode, an open-source TypeScript engine that indexes codebases into a SQLite code graph served to AI agents over MCP: 131k LOC in 3.8s, structural queries under 300ms.
- Led Fleet, an agent evaluation and reinforcement learning platform that recreates applications such as GitHub and Booking.com as interactive UI environments.
- Architected a generative AI platform with model routing, prompt orchestration, tool calling and a custom MCP server for real-time web search; RAG pipelines improved response relevance by 25%.
- Architected a distributed speech-to-image inference pipeline routing ComfyUI workflows across GPU-backed pods through parallel queues.
- Designed and owned the backend of Ivorie, a marketplace serving 5k+ users, with real-time chat handling 10k+ messages/day and Stripe Connect payments.
- Designed an LLM-assisted testing strategy that increased test coverage by 35%.
- Contributed to weather.com, running controlled A/B experiments across video and layout variants.
Feb 2022 – Aug 2022
Software Engineer I, Invest With Tribe
Chandigarh, IN · Remote
- Built a stocks investment platform where experts publish strategies for retail investors, with analytics dashboards and leaderboards used by 1k+ users.
- Implemented scheduled jobs for live price updates and return calculations; integrated AWS SQS and Twilio WhatsApp for real-time trade notifications.

Skills

AI & ML: RAG, Vector databases, LangChain, LangGraph, MCP, LLM evaluation, Agent evaluation, RL environments
Languages: TypeScript, JavaScript, Python, Go, C++
Backend: Node.js, NestJS, Express, REST APIs, WebSockets, Prisma, Event-driven architecture
Data: PostgreSQL, MongoDB, Redis, DynamoDB
Cloud: AWS (EC2, S3, SQS), Docker, CI/CD, Git
Frontend: React, React Native, Expo, Next.js, Tailwind CSS

Recognition

EV Charging App recognised among the Top 30 Startups in India (Inc42, 2022).
Mentored 50+ students in React and React Native development at PrepBytes.

Education

Bachelor of Computer Applications, Techno Main Salt Lake

2020 – 2023 · Kolkata, IN