FlowCV Logo
Education

B.S. in Computer Science

University of California, Santa Barbara
06/2027

College of Creative Studies

GPA: 4.0

Experience
  • •Implemented depthwise convolutions in a Rust-based ML compiler, then made them 16x faster than im2col
  • •Reduced MobileNet compile time from 40 mins to seconds by eliminating redundant middle-end passes
  • •Built cycle estimation scripts to evaluate novel scheduling algorithms, achieving 5.8x faster MobileNetV2
  • •Extended Halide autoscheduler to target In-Memory Compute hardware in the compiler backend
  • •Developed an automated QA pipeline in C and Python to detect faulty eval kits 3x faster
  • •Benchmarked and analyzed energy efficiency on various customer prototypes (ex. Google and Tile)
    • Projects
  • •Beat all other vision-only Geometry Dash models with a Mixture-of-Experts DQN agent
  • •Developed C++ mod to send video buffer and jump controls to the AI model
  • •Implemented and evaluated various RL methods including NoisyNets, Dueling DQN, and DDQN
  • •Augmented Qwen2.5 with programming/web tools using Langchain and PyTorch
  • •Integrated text-to-speech and speech-to-text capabilities with realtime streaming
  • •Benchmarked novel methods to improve LLM performance, such as Attention Sinks and GraphRAG
  • •Trained a multi-task detection/segementation model in PyTorch to track ping pong balls, tables, and players
  • •Integrated SLAM techniques to capture 3D human motion from video
  • •Developed a pipeline in Unity to transform the data into a live 3D reconstruction
    • Skills
    Machine Learning

    PyTorch, ONNX, ML Compiler

    Programming

    Python, C++, TypeScript/Node.js/React, Rust