Hi — I'm Jerome, an independent ML researcher at Algoverse. Previously, I was a graduate student at Université de Montréal and Mila. My current research interests are improving and interpreting reasoning in foundational models and SLMs, with broad interest in fundamental reinforcement learning and representation learning.

I've built ML pipelines for drone flight efficiency at Shearwater Aerospace, improved credit loss forecasting models for regulatory risk pipelines for clients (Citibank NA, PNC Financial Serivices) at TCS, and built computer vision models for fault segmentation and on-device deployment at Ignitarium. My undergraduate was at TKM College of Engineering in CS, focused on MRI medical imaging with deep learning.

I like to sprint on hills, solve chess puzzles, listen to film scores (expansion, philharmonic, love and work) and learn about different cultures and languages.



Efficient Off-Policy Meta-Reinforcement Learning via Probabilistic Context Variables

Read →

Interpreting Latent CoT Reasoning as Dynamical Systems

Shreya Sanjay Boyane, Sabari Iyyappan Duraipandian, Manju Nagesh, Jerome Francis, Archana Vaidheeswaran, Kevin Zhu

ICML 2026 FoGen Workshop, 2026


PerplexMATH: Steering LLMs Toward Mathematical Reasoning

Jerome Francis, Kamesh R, Serena Pei

ICML 2025 NewInML Workshop, 2025


A Novel approach for b value optimization in IVIM imaging using an LSTM based deep learning network on simulated data

Abin Shoby, Jerome Francis, Jini Raju, Ushadevi Amma C., Ansamma John

International Journal of Medical Engineering and Informatics, 2023