Tejas Prabhune

I'm a big fan of machine learning, systems, economics, social science, debate, painting, and table tennis. Trying new things!

  • Sim Francisco — modeling distributionally accurate city simulations. simfrancisco.org
  • Wispr — post-trained and deployed multilingual models. wisprflow.ai
  • Machine Learning @ Berkeley — ex-president (Sp26), ex-VP of Education (Fa25). ml.berkeley.edu
  • CS 189/289 (Intro to Machine Learning) — lead TA twice. Made the homeworks for information theory, transformers, diffusion, and flow matching.
  • Berkeley Speech Group — speech-driven 3D avatars of the vocal tract, brain-computer interfaces for virtual embodiment, and audio-visual dialogue from avatar-based LLMs. Publications, Google Scholar
  • NPDA Debate — NPTE Champion 2026. npdadebate.com, goldenturn.org
  • Research — Contrastive learning-based pretraining improves representation and transferability of diabetic retinopathy classification models. Nature Scientific Reports, 2023. Paper