Berkeley AI Research (BAIR)

Spearheaded research and publications in robot manipulation & model training acceleration with Ken Goldberg.

Developed compression system achieving 70x cost reduction in transferring video data across 21+ research labs.

Proficient in software/hardware stack of 7 bimanual, humanoid, quadruped robots with 200+ experiment hours.

Key skills: VLMs, ROS, neural networks, reinforcement learning, data augmentation, cloud robotics, sim-to-real.

TikTok

Solely deployed five full-stack features optimizing the end-to-end post lifecycle for all 2 billion users worldwide.

Patched major security vulnerability found in 500,000 video uploads in collaboration with multiple safety teams.

Enhanced frame extraction, transcoding, API authentication pipelines, saving ~20 hours/week for a team of 12.

Key skills: video encoding, cloud databases, server infrastructure, live deployment, Agile project management.

GamesCrafters

Supervised 3 team projects in CV, Perception, GPU architecture for a computational game theory research lab.

Designed AR/VR web and mobile app displaying live on-board analysis of best moves, extensible to 100 games.

Reengineered sequential solver algorithm to run vectorized and parallel on cloud GPUs, achieving 30x speedup.

Model Predictive Control Lab

Directed research in reinforcement learning and high-speed (20 mph) autonomous racing on scale-model cars.

Implemented vision-action models & controllers for motion planning and low-visibility (2ft) obstacle avoidance.

Experience in bridging sim-to-real gap, from 3D modeling and OpenAI Gym simulations to physical race tracks.

University of California, Berkeley

GPA: 3.95   |   Honors: Eta Kappa Nu (HKN)

Built and optimized a 5-stage Risc-V processor from scratch in Verilog for ASIC tapeout. 5-month solo project.

Parallelism and pipelining, FPGA & ASIC digital modules, cache & memory design, industrial verification tools.

Expanded sensing and control ability (LiDAR, PID) of self-engineered voice-controlled robot car from EE16B.

Analog ICs, multistage CMOS, semiconductor fabrication, small/large-signal amplifiers, area/power/reliability.

CS 61B: Data Structures and Algorithms
CS 182: Deep Neural Networks
CS 189: Machine Learning
EE 106: Robotic Manipulation and Interaction
CS 152: Computer Architecture
EE 120: Signals and Systems

Robo-DM: Data Management For Large Robot Datasets

Chen K, Fu L, Zhang Y, Goldberg K, et. al.

Available Here.

HANDLOOM 3.0: Interactive Bi-Directional Cable Tracing Amid Clutter

Yu J, Zhang Y, Goldberg K, et. al.

Available Here.

Languages

Python, Go, C, C++, C#, CUDA, VHDL, System Verilog, Java, Swift, Javascript, HTML, CSS, SQL

Software

PyTorch, TensorFlow, ROS, OpenCV, Redis, Docker, AWS, Azure, Google Cloud, Cadence, LT Spice

Hardware

Robotics, GPU, ASIC, FPGA, System on Chip, Design & Verification, Spatial & Cache Architecture

Workflow

Git, Jira, JetBrains, Visual Studio Code, Microsoft Office. APIs: ChatGPT, Claude, Gemini, Llama