Projects
Eighteen active projects across foundation models, applied AI, and systems research. Each one connects fundamental science to a real-world question.
Multilingual foundation models for Indian languages
A 7B-parameter open model family trained on 22 Indian languages for downstream fine-tuning across legal, medical and educational corpora.
Early stroke detection from multi-modal clinical streams
Transformer fusion of EHR, neural imaging and temporal vitals achieving 94% precision on validation cohorts across three partner hospitals.
Sub-millisecond inference for constrained devices
Pruning and distillation framework cutting transformer inference latency by 40% on ARM and RISC-V edge hardware with no accuracy compromise.
Interpretability primitives for deep models
A library of attribution, probing and counterfactual tools for understanding model decisions in clinical and legal applications.
Neuro-symbolic reasoning over knowledge graphs
Hybrid systems combining graph neural networks with symbolic logic for verifiable multi-hop reasoning across structured domains.
Vision-language-action for mobile manipulation
End-to-end policy learning for warehouse and home robots with simulation-to-real transfer on commodity hardware.
Completed projects