A running log of building a language-queryable 3D scene memory system using Gaussian Splatting and CLIP feature fields.
A Next-Best-View (NBV) planning system that autonomously decides optimal camera positions to reconstruct 3D objects.
Real-time volumetric reconstruction using ORB-SLAM3, stereo depth, and TSDF fusion—streamed live from an iPhone.
Global Entry Appointment Finder
A PyTorch CNN trained on the Fruits-360 dataset that classifies fruits in real-time via webcam using OpenCV.
A comprehensive web and SMS-based healthcare communication platform connecting patients and healthcare providers through secure messaging, forums, and voice calling capabilities.
A streamlined implementation of Neo's learned query optimization approach, using tree convolution networks to beat PostgreSQL's cost-based optimizer on complex join queries.
A real-time weed/crop classifier using YOLOv8.
Real-time ASL-to-text translator using MediaPipe, OpenCV, and PyTorch, training a CNN on the ASLLVD dataset and augmented data to interpret the ASL alphabet, expressions, and idiomatic phrases.