Projects
Flex VR Wellness
MIT Reality Hack 2024 - Hardware: Creative Inputs/Outputs Winner
● Collaborated with a talented team of 5 to develop the FlexVR Wellness ecosystem to enable remote electro-stimulation therapy.
● Designed a system where the therapist uses an AR headset to enhance their workflow, communicate with patients, get live data, and control the patient’s electro-stimulation therapy while the patient is in a calm stress-reducing VR environment.
● Wrote firmware and did fabrication + hardware design of the project during the hackathon.
● First developers ever to create a system enabling cross-play between the Qualcomm Snapdragon Spaces AR system and the Meta Quest Ecosystem.
Rewind (Apple Vision Pro WebXR App)
HackMIT 2024 - Intersystems AI Challenge 3rd Place
● Created Rewind, an memory preservation web app that allows users to store, revisit, and share their memories in various formats such as video, photos, text, and audio.
● Designed a system for users to query memories using natural language search.
● Developed pipeline that integrated gaussian splatting to generate 3D scenes from user-uploaded videos for immersive memory viewing.
● Used Three.js and WebXR to allow an Apple Vision Pro user to query and view their 3D memories.
SeamPose PCB
A printed circuit-board I designed for a research paper. Faciliates 8 conductive seams to enable upper-body pose estimation through capacitive sensing. The capacitive signals were sent to a nearby laptop. These signals combined with a vision based ground-truth produced a machine learning model that could predict upper-body pose estimation.
Co-Second Author on the Paper for contributions (Freshman Year)
Realtime Gaussian Splat Localization
● Led the design and creation of a python library to facilitate real-time localization into pre-made gaussian splat scenes and structure from motion (SfM) models based on recent research advancements in feature matching algorithms for 6-DoF visual hierarchical localization.
● Optimized a Pytorch pipeline reducing localization time through pre-loading models and optimizing structure for smaller-scale scenes.
● Converted open-source gaussian splatting software into a form usable by Windows, using CUDA and COLMAP documentation.
● Developed a Flask backend + Three.js frontend of a WebXR based interface for phone and VR headsets to enable synchronous collaboration on hardware projects with interactions powered by 3D-data gathered from localization of images into gaussian splats.
● Camera localization library created was implemented and used in CHI 2025 research paper submission.
Electromagnetic Thor Hammer
Senior year highschool project, originally for class, but ended up being more for fun since it took longer than expected to finish.
● Used two heavy-duty door electromagmets for the electro-magnetism.
● Controlled by Arduino and fingerprint scanner.
● Powered by four 12V lead-acid batteries wired in parallel.
● Donated to my Highschool Digital Logic Design Teacher for his inspiring of me to pursue engineering (his fault that I don't sleep)