Vashisth Tiwari

Hello! I’m a first-year PhD student at Carnegie Mellon University’s Language Technologies Institute.
I completed my M.S. in Electrical and Computer Engineering at CMU and a B.S. in Physics and Mathematics at the University of Rochester. In summer 2025, I worked at Google Research as a Student Researcher with Giulia DeSalvo, focusing on synthetic data generation and reinforcement learning.

My research focuses on making smaller models better and more efficient—through both compression and improvements to data and training. I’m particularly interested in distillation, synthetic data generation, and reinforcement learning in post-training settings, aiming to improve data quality and better understand the dynamics of RL-based fine-tuning. Previously, I’ve explored improvements from the compression and efficiency side, including speculative decoding, model sparsity, and self-speculation for efficient inference.

I’ve been fortunate to collaborate with mentors across physics, math, and machine learning. I’ve worked with Prof. Beidi Chen (CMU) on efficient inference, and with Prof. Emma Strubell (CMU), Lucio Dery (DeepMind), and Clara Na (CMU) on active and model-aware synthetic data generation.

I’m always excited to connect with others in the field, so feel free to reach out if you’d like to chat about research, potential collaborations, or just want to say hi!

Photography portfolio →

When I'm not hunching over my laptop, you'll find me behind a camera lens—check out my photography portfolio. I spent much of my life playing badminton, including on the team at the University of Rochester, and now I'm attempting (and miserably failing) to learn tennis. You might also spot me tossing a frisbee at the CUT or hiking around Pittsburgh.

And yes, I love food and coffee, perhaps a bit too much. I'm always on the lookout for new spots—so if you have recommendations, send them my way!