Summary
Sabyasachi Sahoo is a PhD candidate and visiting researcher at Mila/Université Laval and Inria, focused on making foundation models more reliable, generalizable, and efficient for realistic compute and data budgets. He combines deep research in test-time adaptation, OOD detection, adversarial robustness and continual learning with practical industry experience fine-tuning vision-language and LLM systems at ServiceNow and Amazon. His work spans robotics, medical imaging, and self-driving platforms from prior roles at IISc, NVIDIA and Donut Research Labs, reflecting a strong track record of moving rigorous ML methods into safety-critical applications. Notably, he has developed layer-selection methods for test-time adaptation and used self-distillation of CLIP for zero-shot tasks—approaches that bridge theoretical insights and deployable solutions. Based in Montreal, he is actively publishing and collaborating across academia and industry and welcomes research partnerships.
9 years of coding experience
6 years of employment as a software developer
Doctor of Philosophy - PhD Machine Learning, Doctor of Philosophy - PhD Machine Learning at Mila - Quebec Artificial Intelligence Institute
Master of Technology (M.Tech.) Computational Science, Master of Technology (M.Tech.) Computational Science at Indian Institute of Science (IISc)
Bachelor of Technology (B.Tech.) Mechanical Engineering, Bachelor of Technology (B.Tech.) Mechanical Engineering at National Institute of Technology Surat
English, Hindi, Odia