ML/AI Engineer · MS Computer Science @ Northeastern University (GPA: 3.84)
I build machine learning systems that solve real problems — from spatiotemporal CNNs for live soccer analytics to full-stack AI applications on AWS serverless infrastructure. Currently focused on NLP, computer vision, and deploying ML at scale.
- Context-Aware Toxicity Detection — NLP research using game state features (player deaths, win/loss, performance metrics) alongside chat text to improve toxicity classification in multiplayer games. Fine-tuning transformer models on ToxBuster and CONDA datasets.
- CodeFlow — A VS Code extension for Blueprint-style visual code flow visualization, combining AST parsing (tree-sitter + TypeScript Compiler API) with interactive graph rendering (React Flow + elkjs).
| Project | What It Does | Tech |
|---|---|---|
| Tennis Prediction Model | Predicts ATP match outcomes using 40 years of data with ELO ratings, H2H stats, and rolling metrics. ~70% accuracy with XGBoost. | Python, Scikit-learn, XGBoost |
| MiraAI | AI astrology chatbot with birth chart generation, deployed on AWS serverless (CloudFront, Cognito, API Gateway, Lambda, DynamoDB, Bedrock) with Terraform IaC. | React, AWS, Terraform |
| AR Pong | Augmented reality pong on any flat surface using chessboard pose estimation and real-time hand tracking. | Python, OpenCV, MediaPipe |
| Beaver's Day Out | Sokoban puzzle game with infinite procedurally generated levels, verified solvable via BFS. | Java |
Most recently: AI Engineer @ EasyChamp — built spatiotemporal CNN models for soccer action classification (80.5% CV accuracy, 87% test) and a real-time video analytics pipeline using PyTorch, RF-DETR, SAM, and ONNX.
ML/AI: PyTorch · TensorFlow · OpenCV · Scikit-learn · ONNX · Ultralytics · LangChain · Hugging Face
Cloud & Infra: AWS (Lambda, DynamoDB, Bedrock, Cognito, API Gateway, CloudFront, S3) · Terraform · Docker · Kubernetes
Languages: Python · Java · C++ · JavaScript · SQL
Web: React · Node.js · Flask · Django