I am a software engineer on the ML Model Compiler Team at Nuro.
My work involves accelerate ML model inference speed after training, both from the infra and the modeling perspective.
During college, I researched on quantization with Professor Rayan Saab.
In my free time, I enjoy playing ultimate frisbee!
Mar. 2024 - Present
I am a software engineer on the ML Infra team, working on an ML compiler that optimizes inference latency for models on autonomous vehicles. Currently, I am exploring quantization to accelerate inference on cost-effective GPUs optimized for low-precision calculations.
Nov. 2023 - Mar. 2024
I was a software engineer on the payment onboarding team, focusing on the merchant registration process to integrate them into the Square ecosystem. I collaborated with product managers, designers, and engineers to develop new features for the Canada registration process, ensuring compliance with CA KYB requirements.
Sept. 2022 - Nov. 2021
As part of the Stripe's privacy infrastructure team, I designed and implemented an ML-based workflow to automatically infer metadata within Stripe's data ecosystem, integrating it with Trino for attribute-based access control and enabling data obfuscation during queries. Additionally, I helped develop a data annotation system and collaborated with cross-functional teams to build tools that facilitated its adoption.
June 2021 - Sept. 2021
As a Software Engineer Intern on the Simulacra project, I redesigned a distributed system to enhance scalability, which significantly reduced user response time. I optimized the naive K-means algorithm for the 'rendering items on map' feature, achieving a 1,000x speed improvement and creating a smoother user experience.
Apr. 2019 - Dec. 2021
I worked in the CSE department at UCSD under Professor Gary Gillespie. I worked as an undergraduate TA for CSE 12 (Basic Data Structures) and CSE 15L (Software Tools). I was the head tutor for CSE 12 from Winter 2021.
Sept. 2021 - June 2022
I worked with Professor Rayan Saab and his PhD Jinjie Zhang on developing a greedy path-following algorithm that performs quantization on pre-trained ML models with provable error bounds. I implemented the algorithm in PyTorch. I conducted experiments using AWS to quantize state-of-the-art neural networks (ResNets, VGGs, etc.) to verify the effectiveness of the algorithm.
Oct. 2020 - June 2022
I was an undergraduate researcher in the Lim[b]itless group at Qualcomm Insitute working building a pipeline that increase the availability of prosthesis to the amputees. I worked on using transfer learning over the DeepLab deep learning network to perform background removal on the input images.
Jan. 2021 - June 2022
I completed my honor's project under Professor Mareike Dressler. I investigated on the numerical effects that different polynomial bases have when trying to certificate the nonegativity of polynomials. My thesis can be viewed here.
yixzzhou [at] gmail [dot] com
37.3688, -122.0363