Vashisth Tiwari

PhD @ Carnegie Mellon

prof_pic_2.jpg

Hi! 👋

I’m a first-year PhD student at Carnegie Mellon University’s Language Technologies Institute, advised by Emma Strubell and Zico Kolter. Previously I completed my masters at CMU and a bachelor’s in Physics and Mathematics at the University of Rochester.

My research focuses on developing better, more efficient LLMs by improving training dynamics and data quality. I am particularly interested in knowledge distillation, synthetic data generation, and reinforcement learning for post-training.

I’ve been fortunate to learn from wonderful mentors, including Giulia DeSalvo at DeepMind; Beidi Chen, Lucio Dery, Clara Na at CMU.


Outside of work, I love photography—check out my portfolio and Instagram.

I'm always on the lookout for new food and coffee spots—if you have recommendations, send them my way!


news

May 19, 2025 Student Researcher @GoogleResearch (📍Seattle) w/ Giulia DeSalvo on synthetic data for post-training
May 15, 2025 ACL Main (Energy considerations on LLM inference)
Apr 14, 2025 Will be joining CMU-LTI PhD Fall’25 cohort
Jan 23, 2025 ICLR'25 (MagicDec)
Aug 19, 2024 Released MagicDec: Breaking Throughput-Latency Trade-off for Long Context Generation with Speculative Decoding. Blog