click for more
click for more
Developed an app that allows users to chat with a real-time voice bot that gathers domain-specific knowledge to host expert-level conversations on any topic and can be used for anything including mock interviews, small talk practice, language learning, pitch rehearsals, therapy sessions, and customer service simulations
Trained a convolutional autoencoder using Tensorflow to transform video frames into a latent space and used a vector sampling and interpolation technique to synthesize frames for a new video using a GAN to enhance synthetic video realism. Won the grand prize at UC Berkeley Datathon 5.0!
App combining LLMs and stable diffusion (before the release of Dall-E and big commerical models) to generate WikiHow-style instruction pages with AI-generated images, crafted with FastAPI, Next.js, and Jinja2 to embed AI-generated images alongside text instructions on each page.
C script utilizing Intel Intrinsics SIMD instructions, OpenMP for multithreading, and OpenMPI for multi-processing to apply convolutions for video processing to blur, sharpen, or apply other effects to videos
Java-based 2D game where you search for a goals in the dark on randomly generated maps dodging enemy encounters where you play rock-paper-scissors life points on the line, try and climb as many floors as you can!
This was my first ever HTML+CSS+Javascript mini-project! Thought it was cute so I'll leave it here, what a long way we've come :-)
Kdramas I've watched so far, send recs!
Hover for ratings...
Created entirely by Nathan Shum using React.js + Next.js