Sanket Mhatre
Researcher at Proximal.
working on evals. before this, i was at Microsoft Research — helped GitHub evaluate their coding agents and trained models to understand excel formulas.
i led the gen ai team at PocketFM after starting there as a backend developer. before that, i interned at Trell and a few other early-stage startups.
i'm also trying to find my way into ai safety research.
experience
- jul 2026 — now MTS | Proximal
- dec 2024 — jul 2026 Research Fellow | Microsoft Research
- sep 2022 — nov 2024 SDE-1 → SDE-2 | PocketFM
- oct 2021 — apr 2022 SDE Intern | Trell
- aug 2021 — oct 2021 Backend Developer Intern | Zhiffy
publications
- FrontierSWE v2: Evaluating AI on Ultra-Long Horizon Software Engineering Tasks [blog]
Rishyanth Kondra*, Sanket Mhatre*, Akshit Kumar*, et al.
Proximal, 2026
- UnWEIRDing LLM Entity Recommendations [arxiv]
Aayush Kumar*, Sanket Mhatre*
AAAI 2026, LM4UC Workshop
- SWE-Sharp-Bench: A Reproducible Benchmark for C# Software Engineering Tasks [ieee]
Sanket Mhatre*, Yasharth Bajpai*, Sumit Gulwani, Emerson Murphy-Hill, Gustavo Soares
Best Paper Award, AIWare 2025
- A Hybrid Method for Fake News Detection using Cosine Similarity Scores [ieee]
S. Mhatre, A. Masurkar
ICCICT 2021