Soham Govande

Soham Govande is a driven student at Stanford University, pursuing both a BS in Computer Science with a focus on AI and an MS in Computer Science specializing in systems. His academic trajectory is heavily influenced by his interest in the intersection of machine learning and systems, which has led him to explore hardware-aware model optimization. sohamgovande+1

His professional experience includes a significant role as a Member of Technical Staff at OpenAI, starting in 2025, following a tenure as a Deep Learning Algorithm Engineer at NVIDIA in the same year. He also gained valuable research experience as an Undergraduate Research Assistant at the Stanford Artificial Intelligence Laboratory (SAIL). sohamgovande+2

Soham is an active contributor to the field of AI research, notably co-authoring the paper "Chipmunk: Training-Free Acceleration of Diffusion Transformers with Dynamic Column-Sparse Deltas." This work introduces an innovative method to significantly speed up diffusion transformers for both video and image generation, demonstrating his expertise in optimizing machine learning models for hardware efficiency. sohamgovande+3

Beyond research, Soham has a strong entrepreneurial spirit and a knack for developing practical applications. He co-founded AdmitYogi, a platform designed to democratize access to college admissions guidance, and developed HideGPT, a tool to mask AI-generated text. His diverse projects also include Cardr, a debate evidence collector, and contributions to hackathon-winning projects like ShipSense and PodPlex. stanforddaily+5

His technical skills are broad, encompassing React, Next.js, and Chakra, as evidenced by his Devpost profile. He demonstrates a consistent interest in Machine Learning/AI, reinforced by his academic pursuits and project portfolio. devpost