Software Product Engineering student in Pune. I spend most of my time in other people's codebases — type checkers, protocol stacks, SDK generators and hardware description frameworks — finding a bug, proving it, and sending the fix.
| Project | Org | What it is | What I did there |
|---|---|---|---|
| pyright | Microsoft | Python type checker | Enum tuple values, class-pattern narrowing, abstract-class instantiation, overload selection |
| kiota | Microsoft | API client generator | Code-generation correctness across seven target languages — name collisions, serialization, imports, enum handling |
| typespec | Microsoft | API description language | OpenAPI 3 emitter — duplicate schemas, SSE event constraints |
| quiche | Cloudflare | QUIC / HTTP3 | RFC 9002 congestion control and loss recovery compliance |
| pingora | Cloudflare | Rust proxy framework | HTTP/1 upgrade handling |
| rohd / rohd-hcl | Intel | Hardware description in Dart | Clock generation, CSR access rules, arbiters, namespace collisions |
| auto-round | Intel | LLM quantization | Autocast handling, device placement, padding masks |
Every one of those started the same way: read the code until the bug is reproducible, write the test that fails, then fix it.
HireOnyx — an AI-powered job portal. MERN, with the matching layer doing the interesting work.