Hi, I’m Vasu Sharma 🙂

I build multimodal foundation models that power generative AI agents. Currently, I lead applied research at Meta AI (FAIR), focusing on 100B‑parameter LLMs and self‑supervised vision & audio‑video systems.

Get in touch

Career

Applied Research Scientist Lead — Meta AI (FAIR)

Aug 2022 – Present · Menlo Park, CA

Tech‑lead for Llama 3/4, Chameleon and MoE initiatives.
Designed MaVIL, a SOTA masked audio‑video learner.
Developed text‑quality scoring that halved training corpus size with no loss.

Applied Scientist — Amazon Alexa AI

Aug 2021 – Aug 2022 · Sunnyvale, CA

Co‑founded Alexa Arena embodied‑AI benchmark; built real‑time multimodal transformers.

Quantitative Research Analyst — Citadel GQS

Aug 2019 – Aug 2021 · Chicago, IL

Automated cross‑asset alpha construction with large‑scale ML pipelines.

Education

M.S. Language Technologies — Carnegie Mellon University

GPA 4.19/4.33 (dept rank #1)

B.Tech. CSE — IIT Kanpur

CPI 9.9/10; All‑India JEE Rank 165

AISSCE — St. Columba’s School

97 %

Selected Publications

DINOv2: Learning Robust Visual Features without Supervision — TMLR 2024
MAViL: Masked Audio‑Video Learners — NeurIPS 2023
Chameleon: Mixed‑Modal Early‑Fusion Foundation Models — arXiv 2024
Alexa Arena: Embodied‑AI Platform — NeurIPS 2023
FLAP: Fast Language‑Audio Pre‑training — ASRU 2024

See everything on Google Scholar ↗.

Contact

I’m always open to research collaborations and speaking engagements.

Email me LinkedIn Scholar Resume PDF