Projects
Work spanning machine learning, computer vision, spatial computing, and systems.
- 1.Stylescape: Real-Time Stylized Passthrough on Meta Quest 3
XR / Computer Vision · 2026
A real-time stylized passthrough pipeline for the Meta Quest 3 that decouples neural style-transfer inference from the display rate via depth-based reprojection, holding 60+ fps fully on-device with no PC tether or cloud. Adds an auto-stride controller and high-pass detail injection to recover sharpness at low input resolution.
- 2.Power Lever: GPU-Efficient LLM Inference Gateway
Winner, Stanford Hackathon · 2026
An inference gateway that dynamically routes LLM prompts to right-sized GPU hardware across 4 tiers, cutting energy use by 75% on simple queries while reserving high-end GPUs for complex tasks.
- 3.DYNAMO: Reinforcement Learning Portfolio Manager
Reinforcement Learning · 2026
An RL agent that manages a portfolio across 10 asset classes, achieving 15.76% annualized returns at a 1.44 Sharpe ratio and outperforming traditional strategies by 38–54%.
- 4.Eous: Embodied AR Robot Assistant
Winner, Stanford Hackathon · 2026
A hands-free AR system pairing AR glasses, a smartphone, and a Raspberry Pi robot for gesture and voice control with live camera feedback — all processed on-device with no cloud dependencies.
- 5.VR Healthcare Training Simulation
Winner, MIT Hackathon · 2024
A VR healthcare simulation on the Apple Vision Pro for CPR and first-aid training, pioneering SharePlay integration for collaborative learning with iPhone users. Showcased to The Venture Reality Fund, where the concept was carried forward.
- 6.AgenTeX: Image-to-LaTeX Agent
Winner, AgentOps Hackathon · 2025
An AI agent built with OpenAI’s Agents SDK that converts handwritten math to LaTeX with 91% accuracy. Featured as an AgentOps hackathon winner with a board-member recommendation for commercial launch.
- 7.Medical MRI Image Reconstruction
Deep Learning Research · 2024
AI-based models for medical MRI image reconstruction that cut required scan times by 55–70% while preserving image quality, reducing reconstruction error by ~60% relative to baseline deep learning models.
- 8.Supreme Court Case Prediction
1st Place, Stanford CS109 · 2025
A Bayesian case-prediction framework achieving 73% top-3 accuracy across 11 possible Supreme Court outcomes from observable case attributes. Featured as a winning example project in Stanford’s CS109.
- 9.SynchroSound: Facial-Expression Song Selection
iOS / ML · 2024
An iOS app that reads facial expressions to recommend mood-matching songs, integrating SwiftUI, UIKit, and SwiftData with Google Cloud Vision and Spotify’s Web API. Validated across 500+ test images, improving mood-alignment accuracy by 82%.
- 10.Music Recommendation System
Machine Learning · 2025
A recommendation system using SVD and PCA over 60+ audio features from 7,000 Spotify songs, uncovering patterns beyond traditional metadata and projecting features to identify similarities across 35 feature combinations.
- 11.
- 12.
- 13.Neural Housing Predictor
Academic Project · 2024
A machine learning model achieving a 35% accuracy improvement in property valuation through advanced data processing and neural-network architecture.
- 14.LZW File Compression
Systems Programming · 2024
A high-performance file-compression implementation reducing file size by ~50% relative to standard methods, enabling faster and more efficient transfers.
- 15.Heap Allocator
Systems Programming · 2024
A custom memory-management system implementing malloc, realloc, and free in C, improving memory utilization by 20% across 100,000 requests.
- 16.Custom Unix Shell
Operating Systems · 2024
A command-line Unix shell supporting file manipulation, redirection, and running 10+ programs concurrently, tested across 100+ real-world scenarios.