Hi I'm David, I'm an undergrad at UC Berkeley, software engineer, ML researcher, and overall problem-solver who enjoys bridging cutting-edge research with real-world engineering problems. With a background in research and software engineering, I want to create innovative and scalable software that incorporates distributed computing and machine learning into the latest engineering challenges.
Seeking an internship where I can apply my industry and research experience to engineer modern solutions
BS in Electrical Engineering and Computer Science, Minor in Mathematics
Software Engineer Intern
Focused on Data Engineering and Large Language Models (LLMs) resource optimization. Enhanced resource utilization and runtime efficiency of large language models and distributed computing systems by 31% and up to 72% in specific cases through targeted algorithm optimization research
Undergraduate Researcher
Improving the scalability of Gaussian Splatting technology for NeRF research, resulting in significant performance enhancements and the development of a novel class-agnostic algorithm
Software Engineer Intern
Engineered a responsive and scalable web application with Firebase, Tailwind, and React that streamlined inventory and transactions for the UC Berkeley community of 40,000+
Software Engineer Intern
Integrated conversational AI voice assistant technology for diverse clientele applications, orchestrating the development of scalable, secure systems utilizing JavaScript, React, and Flutter frameworks
CS 180 is an undergraduate computational photography and computer vision class at UC Berkeley. I implemented fundamental image processing algorithms entirely from scratch in Python, beginning with basic physics and optics principles of image formation and advancing to modern deep learning architectures for visual understanding. Through intensive programming projects, I built a comprehensive toolkit spanning classical techniques like image warping and stereo vision to cutting-edge applications in neural networks and Diffusion models, with each component crafted from first principles rather than relying on existing libraries. The hands-on curriculum bridged the theoretical foundations of computational photography with contemporary machine learning approaches, challenging me to develop sophisticated computer vision systems that demonstrate mastery of both the underlying mathematical concepts and their practical implementation.
This project implements a series of image manipulation techniques such as color manipulation, bilinear interpolation, and pyramid blending for image resizing and morphing. The work showcases the development of a visual experience by applying these transformations to create seamless image transitions and hybrid images.
This project explores image processing techniques, including the application of finite difference operators to compute image gradients and Gaussian filters for edge detection. Additionally, it demonstrates the creation of hybrid images by combining high and low-frequency components from different images, along with advanced blending techniques such as multiresolution blending and Fourier transformations to enhance image quality.
In this third project we create an algorithm which will seamlessly transition from one image to another, or even more! This will involve implementing a correspondence program in order to divide our source images into triangular regions.
This project focuses on creating an efficient and accurate panorama stitching technique by combining multiple images into one seamless panorama. Using advanced image processing methods, it addresses blending challenges, such as dark borders, to achieve smooth transitions at the edges for a more realistic final result.
Creating and Experimenting with Diffusion Models on Pytorch. In this project, we implement UNet denoising models and different types of image diffusion models.
Implemented the NeRF paper, and implemented some changes to the architecture to improve learning capabilities.
Developed a HackMIT-winning project that visualizes cancer progression in MRI scans by parsing 2D slices into a custom NeRF model, segmenting tumors with SAM, and generating 3D meshed visualizations for real-time analysis on a React-based web platform.
Evaluated current adversarial ML techniques for images to determine features and architectural choices that make networks susceptible. Used findings to develop an improved adversarial algorithm with less perceivable noise.
Implemented a k-Nearest Neighbors (k-NN) algorithm to predict geographic coordinates from image embeddings encoded using OpenAI’s CLIP model, achieving a 93% accuracy rate by using mean displacement error (MDE) for evaluation.
Designed and implemented an RNN-based model in Python (NumPy) for language identification of variable-length strings. Implemented parallelized activation layers, linear modules, and backpropagation algorithms.
Engineered an offline voice-control system through signal processing, PCA, and machine learning models. Optimized a closed-loop feedback power system for steady vehicle control.