how i got here
a fairly non-linear flight path.
research labs, a couple of startups, a lot of gpus. most recent first. the full version, funny bits included, lives on linkedin.
01 path
- feb 2026–nowfounding engineer · brief (a16z speedrun)the product-context layer for ai coding agents. san francisco.
- dec 2025–feb 2026applied ai engineer · aira technologiesagents: self-improving, reasoning, evaluation. san francisco.
- jan–aug 2025applied ai intern · aira technologiesreal-time anomaly detection across 6,000+ at&t sites; multi-agent root-cause analysis (5h → 3min). saratoga, ca.
- sep 2024–jan 2025graduate researcher · umass amherst (kdl)mixture-of-experts models for causal effect estimation, with prof. david jensen.
- jan–jul 2024nlp research intern · microsoftselective language modeling for efficient llm pretraining: 10× faster convergence, +6.8% across 15 benchmarks. bengaluru.
- jun–sep 2023
- summer 2022research intern · harvardai labs.
- jun 2022–apr 2023research assistant · iiit sri citydeep learning for healthcare; publications incl. eaai.
- jul 2022–mar 2023research assistant · iit (bhu) varanasiroadsdnet, road-boundary detection; −25% navigation error (acm cv 2025).
- may 2022–jan 2023r&d intern · protohubs.iotext-to-3d generation with gaussian splatting, alongside jagatheeswaran senthilvelan (now head of innovation at noise). remote.
- 2024–dec 2026ms, computer science · umass amhersthealthcare ml, causal inference. amherst, ma.
- 2020–2024b.tech, computer science · vellore institute of technology