Senior Research Scientist on the Surreal team at Meta Reality Labs, working on multimodal agents for smart glasses, egocentric and long-video understanding, and vision-language research.

My work focuses on first-person video, long-context understanding, multimodal post-training, and research systems for real-time contextual assistance. More at theshadow29.github.io.

• Researching multimodal agents for smart glasses, with a focus on egocentric and long-video understanding.
• Building multimodal models and post-training setups for instruction following, reasoning, and real-time contextual assistance.
• Working across the full research stack: benchmark design, distributed training, evaluation, low-latency inference, and deployment-facing integration.
• I enjoy reading the latest papers and articles on LLMs, VLMs, instruction following, reasoning, and agentic workflows.

Website · Google Scholar · CV · LinkedIn · Twitter/X · Email