Build — systems with clean, explainable internals.
Research — course projects that end as papers, not just demos.
Ship — full-stack products, from schema to UI.
Contribute — patches to the Python libraries I use every day.
A fault-tolerant, linearizable key-value store on a from-scratch Raft implementation. 500 seeded chaos runs — partitions, crashes, dropped messages — checked with Porcupine, zero failures.
Go Raft gRPC Distributed systems
A Git-inspired relational database engine. Branch, diff and time-travel your data with AS OF queries.
Python Databases Version control
Distributed background jobs with priorities, delayed execution, retries and failure recovery. Redis + PostgreSQL, Dockerized workers, live monitoring.
Python Redis PostgreSQL Docker
Profile-guided adaptive register allocation that skips graph coloring: stochastic dynamic programming scores each value, and a game-theoretic auction settles who gets the register.
Compilers Dynamic programming Game theory
Useful things, built to be used.
- KernelLens — Root-cause analysis for Linux kernel logs: parsing, an ML anomaly filter, event correlation, then an LLM explanation.
- CodeAtlas — Ask any GitHub repo questions in plain English and get answers grounded in its code.
- ResearchMate — A six-agent pipeline that searches papers, reads them, and drafts a review with cited research gaps.
- Parchi — Turns orders sent by chat, voice note or photo into invoices, in seven languages. Live demo →
- TravelCraft — A trip planner that optimizes which places to see, which day, and in what order around opening hours. Live →
I send patches to libraries I rely on: REA · httpx · boltons · jsonschema · marshmallow · tenacity · funcy
Have a project in mind, or just want to say hello?
I'm a student, open to interesting work — internships, collaborations and open source included.