I'm a CS undergrad exploring machine learning, computer vision, and full-stack development. I learn best by building things myself — sometimes that means starting from scratch, breaking something, and spending way too long figuring out what went wrong. Most of my projects are around AI/ML, from models and RAG systems to the applications built around them. I'm interested in the whole process, not just getting a model to work, but making something usable out of it. Currently learning, building, and figuring things out as I go.

role: CS undergrad, 3rd year
domain: ML / CV / Agentic AI
mode: building > theorising
current_obsession: making agents
                    actually useful
status: always mid-refactor

orbitscreen — Go · WebAssembly · three.js · live demo
Every satellite in orbit (~21k objects), moving live on a 3D Earth, plus every close approach between them over the next 7 days. SGP4 written from scratch in Go and validated against CelesTrak SOCRATES.

satnogs-pipeline — Go · RabbitMQ · PostgreSQL · Kubernetes · live demo
Are small satellites' batteries OK? Decodes volunteer radio telemetry from SatNOGS and checks every reading against sunlight and the Earth's shadow, on a live 3D status page.

Morrow — Whisper · Groq · FastAPI · React · live demo
AI meeting assistant: upload a recording and get the summary, decisions, action items and open questions, then chat with the meeting. Answers only from what was actually said.

CODEX — LangChain · Tavily · FastAPI · React · live demo
Multi-agent research assistant: a search agent finds sources and a writer agent turns them into a cited report. Also has a PDF Q&A mode that answers with page numbers.