Formal Methods · Software Systems · Machine-Assisted Reasoning
I am a recent Computer Science and Engineering graduate from the Bangladesh University of Engineering and Technology (BUET) and currently a research intern at the University of Illinois Urbana-Champaign.
My primary current work with TLAPS-Bench examines how AI systems can construct and complete machine-checkable TLA+ proofs, and how such systems should be evaluated rigorously. I also contribute to SREGym, which evaluates AI agents on realistic Site Reliability Engineering problems in live system environments. More broadly, I am interested in formal verification, machine-assisted reasoning, dependable distributed systems, and software reliability.
- Machine-assisted formal reasoning — I contribute to TLAPS-Bench, a benchmark for evaluating AI systems on completing and constructing machine-checkable TLA+ proofs.
- AI agents for software reliability — I contribute to SREGym, an AI-native platform for developing and evaluating SRE agents on realistic cloud-system failures. My work includes reproducible Kubernetes incidents that test diagnosis and safe recovery under degraded observability.
- Bengali speech representations — My undergraduate research studied where Bengali phone-like information emerges across Whisper encoder layers using speaker-disjoint probing and cross-model comparison.
Layer-wise Probing of Whisper's Encoder Representations for Bengali Phone-like Units
Munim Thahmid, Sadia Sharmin
Accepted to INTERSPEECH 2026.
Before focusing primarily on research, I spent over a year at Yobo AI building backend systems and testing infrastructure for production AI voice-agent applications.
- Formal methods: TLA+, TLAPS / TLAPM
- Programming: Python, C/C++, TypeScript / JavaScript, Bash
- Systems: Linux, Docker, Kubernetes, Git
- Research and ML: PyTorch, Whisper, XLS-R, scikit-learn, pandas, NumPy