Selected work

More Projects

01

Lattice

Finalist — Hack the North 2025Grand Prize Finalist — Stanford Treehacks 2026

Distributed multi-sensor volumetric capture leveraging multi-Kinect LiDAR arrays with Procrustes and ICP-based spatial alignment, zero-copy ring buffers for lock-free inter-process communication, and 4DGS-style reconstruction to generate navigable 3D point clouds with Git-style holographic replay across desktop, web, and AR/VR

C/C++C#.NETPythonHTML/CSSOpenCVUnityKinect SDK
02

DataBot

Finalist — LAHacks 2025

Tool for analyzing slow or inefficient SQL queries; uses Google Gemini to suggest optimized rewrites, explain query plans, and surface index recommendations. Backend is built in Go for high throughput.

GoTypeScriptNext.jsGoogle Gemini
03

SafeGrounds

Platform empowering investors and developers to identify and respect Indigenous lands through transparent, actionable insights. Features interactive mapping with Leaflet, community reporting, contractor response systems, and ML-powered land assessment to prevent encroachment on sacred lands and cultural heritage sites.

TypeScriptPythonFlaskSQLiteNext.jsLeafletAWSChatGPT-4 API
04

Hello World!

1st Place — UTS BluesHacks 2024

Two-way ASL/English translation app using TensorFlow and Google Cloud API to recognize ASL hand gestures in real time and translate them to text (or vice versa, with animated output) and synthesized audio output, making computers more accessible for the deaf and hard-of-hearing community.

JavaScriptTypeScriptTensorFlowGoogle Cloud API
05

Project Klean

Enterprise solution for digital pollution via AI-powered code optimization. Analyzes and optimizes Python code to improve efficiency and sustainability using CodeBERT fine-tuning to identify memory-intensive operations, redundant computations, and suboptimal data structures. Leverages CodeCarbon for sustainability tracking to reduce enterprise-level energy consumption.

PythonPandasMLCodeBERTCodeCarbonStreamlit
06

Bergster

3rd Place — Guelph GDSC Hacks 2024

Cognitive training tool that uses real-time facial recognition and emotion detection to help people with social disabilities train and improve their emotional recognition and expression. Features offline-first design, performance analysis with visual charts, and emotion testing for measurable progress tracking.

ReactNext.jsJavaScriptTensorFlowface-api.jsTailwindCSS
07

Echo

AI-powered memory recall assistant using computer vision and natural language processing to record and process daily activities for users with memory loss, Alzheimer's, and dementia. Uses CLIP for image understanding, OpenAI Whisper for transcription, and Cohere for semantic analysis to help users reconnect with their memories.

PythonJavaScriptNext.jsOpenCVCLIPCohereOpenAI WhisperGoogle Cloud
08

SkinSight

A dermatology prototype that classifies skin lesion images across multiple diagnostic categories using a CNN trained on the HAM10000 dataset. Built with a Flask backend and a lightweight web interface.

PythonFlaskPyTorchTensorFlowMachine Learning
09

Verbate

Best Use of AI — IEEE MakeUC 2023

Built at IEEE MakeUC 2023. An AI interview coach that generates role-specific questions, evaluates user responses in real time using GPT-4, and delivers audio feedback via AWS Polly.

PythonGPT-4AWS PollyNext.js
10

HandPilot

1st Place — YRHacks 2023

Uses MediaPipe's hand landmark detection combined with OpenCV to translate hand gestures into system-level mouse movement, clicks, and keyboard input — allowing hands-free computer control.

PythonOpenCVMediaPipe