CS at BITS Pilani. I build systems, then build the harnesses that break them.
41 PRs merged upstream across 11 organisations, including GCC, Microsoft, PyTorch, NVIDIA, Hugging Face and CNCF.
- A patch of mine is in GCC master. gccrs#4731 went upstream in a maintainer's sync, authorship kept.
- 14 of those 41 fix bugs I found and reported myself, not tickets off a board. One was in my own merged code.
- I found nondeterminism in vLLM's deterministic mode (vllm#51187): repeats match only if every token is reduced at the same RMSNorm block width.
- Sole author of a COLM 2026 workshop paper. When Self-Consistency Backfires: majority voting lowers accuracy on 56–66% of GPQA Diamond problems.
| Project | Merged |
|---|---|
| microsoft/PyRIT | 14 · the CodeAttack converter, plus fixes to TAP/PAIR, scale scorers and dataset loaders |
| gccrs · GCC Rust frontend | 8 · match arm guards, left-to-right argument evaluation, a dead-code fix now in GCC master |
| PyTorch · executorch, torchtitan | 6 · Inspector crashes on non-finite output, wandb tag splitting, scheduler range checks |
| jaeger-ui · CNCF | 4 · GenAI span classification, media rendering, RFC wire-format corrections |
| NVIDIA/garak | 1 · fixed Bedrock scans that failed for every Claude 4.x user |
| huggingface/OpenEnv | 1 · SSRF-safe URL parsing |
| dottxt-ai/outlines | 1 · RFC 4291 IPv6 structured-output type |
| Also | llm-compressor ×2, AI Village ×2, openkruise/agents, openyurtio/raven |
| thrice | Reproduces a reported bug three times in parallel browsers. Precision 0.80, recall 0.80, every attempt published including the two it got wrong. |
| Lockstep | An inference engine and a certifier that checks output stays bit-identical however requests are batched. |
| Shadowbook | A limit order book in Rust. 416ns p50 insert, 100M fuzzed ops against an oracle, zero divergences. |
| Flint | An x86-64 kernel in Rust. Ring 3 isolation proven by a breakout harness. |
| Blink | Your own copy of an open source app, running in five seconds. Gone in ten minutes. Live |
More, including Starling, CAIRN and MIRR, at utkarshbahuguna.me.
I publish what the measurements say. When all-pairs comparison showed half of a Lockstep finding was an artifact, the retraction went next to the result.
Rust C++ Python CUDA/Triton Go Java TypeScript
Portfolio ·
LinkedIn ·
Email



