Projects

Class projects and independent work outside my main research agenda: build-to-learn explorations, course projects, and things I’m proud of that don’t fit neatly into a lab context. Research details some of my past research.

Computer Vision & Computational Photography

Six projects from CS 180 (UC Berkeley, Fall 2024), each reproducing results from a key paper or method.

  • NeRF novel view synthesis render
    Neural Radiance Fields & Facial Keypoint Detection
    CS 180 Final Project · Fall 2024
    Implemented NeRF from scratch (positional encoding, volume rendering, coarse-to-fine sampling) and a convolutional facial keypoint detector trained on the IMM Face Database.
  • Visual Anagram
    The Power of Diffusion Models
    CS 180 Project 5 · Fall 2024
    Explored denoising diffusion probabilistic models: sampling, DDIM, classifier-free guidance, inpainting, and training a small UNet on MNIST.
  • Doe at Night
    Image Warping, Mosaicing & Auto-Stitching
    CS 180 Project 4 · Fall 2024 · Class Choice Award
    Homography estimation, image rectification, and panorama stitching via manual and automatic feature-point correspondence. Implemented Harris corner detection, ANMS, feature descriptor extraction, RANSAC, and the full auto-stitching pipeline.
  • Face morphing
    Face Morphing & Population Statistics
    CS 180 Project 3 · Fall 2024
    Triangulation-based face morphing, computation of mean faces from the Danes dataset, and caricature generation via extrapolation.
  • Barbenheimer
    Fun with Filters & Frequencies
    CS 180 Project 2 · Fall 2024 · Class Choice Award
    Gaussian and Laplacian stacks, hybrid images, and multiresolution blending.
  • Colorized Prokudin-Gorskii glass-plate negative
    Colorizing the Prokudin-Gorskii Photo Collection
    CS 180 Project 1 · Fall 2024
    Automatic alignment of three-channel glass-plate negatives using NCC/SSD image metrics and image pyramid search.

Hardware & Architecture

  • Apple-Sponsored RISC-V CPU Design Contest (1st Place)
    1st of 18 teams · Verilog RTL · Xilinx Zynq-7000 · Spring 2025 (EECS 151)
    Awarded 1st place in a Spring 2025 Apple-sponsored CPU design contest with teammate Neel Gajare. Achieved the contest’s best PPA metrics over two years: lowest resource utilization, 1.06 CPI, and 125 MHz clock frequency. Led microarchitecture and engineering workflow (Git feature-branching, regression testing) to design a fully bypassed, 5-stage RISC-V pipeline in Verilog RTL, including speculative execution via a branch target buffer. Synthesized on the PL fabric of a Xilinx Zynq-7000 (XC7Z020) and validated with a custom SystemVerilog verification suite.