Building Tools That Solve Real Problems
I'm a CSE student at BEU Patna and an AI engineer. I believe in learning by building — not just consuming tutorials — and sharing everything publicly through open source.
My core interest lies in designing systems, not just models — retrieval, pipelines and deployment. I run OpenAgentHQ as the home for that work: agentic systems, LLM infrastructure and developer tooling. I'm currently looking for an AI/ML or GenAI internship.
Real-world tools that solve problems people actually have.
By building — not just consuming tutorials.
Code, notes and lessons, all public and reviewable.
Started a computer science degree and began learning by building, not just following tutorials.
Shipped run-git, a Git automation CLI — and learned how much of the work is everything around the code.
Published OpenAgent Eval and ModelDock under OpenAgentHQ: evaluation for RAG and agents, and a manager for local LLMs.
Released RepoHeart, a multi-agent system that triages issues, reviews PRs and repairs CI from inside GitHub Actions.
Released JEV Model Router, a local proxy that picks the right Claude model for every turn in Claude Code.
RAG pipelines end to end — chunking, embeddings, vector search and the evaluation harness that tells you whether any of it actually improved.
LangGraph workflows and MCP servers: the protocol plumbing that lets models call real tools reliably rather than in a demo.
CLIs that live in a terminal and get used daily — local-first, offline-capable, and fast enough that nobody reaches for the old command.
The topics my recent projects keep pulling me deeper into:
Work merged into other people's codebases — where the review bar is set by someone else.
Anthropic
Databricks
Simplilearn (SkillUp) — powered by Google Cloud
Internshala Trainings
I make short educational tech content for CS students, junior developers and AI beginners — the explanations I wish I'd had when I started. If you're building something in AI, open source or developer tooling, I'd like to hear about it.