Real-time multimodal AI on wearable devices: led the live vision-language-audio assistant demonstration at the DARPA PTG program review at MIT (Oct 2023), on HoloLens and a helmet-mounted Mevo camera-and-microphone rig; evaluated by MIT Lincoln Laboratory.
RL post-training: GRPO in verl (vLLM rollouts, FSDP2, LoRA, custom reward server) for an 8B vision-language model on multi-GPU H100 clusters; reward-structure analysis.
Benchmarks and medical AI: multimodal reasoning evaluation (VERIFY, COLM 2026; MMPerspective, NeurIPS 2025); radiology-report agents (SPIE Emerging Topics in AI 2026); causal time series for fMRI (NeuroImage 2025).
Research in the Department of Computer Science (multimodal lab of Chenliang Xu) and with Axel Wismüller’s imaging group. University listings: ECE graduate students, Wismüller Lab; NRT PhD trainees 2021-22, advised by Chenliang Xu.