Undergraduate at Rutgers double-majoring in Data Science and Computer Science. I build and deploy full-stack ML systems—from RAG research assistants and automated trading engines to VR digital twins and applied ML pipelines—and care about turning data into tools people actually use.
“Well begun is half done.”
Location
Plainsboro, NJ
Education
Rutgers '27
Major
Data Science & CS
Projects
The archive
Aug 2026 – Present
Automated Prediction Market Trading Bot
Python trading system for Kalshi's prediction markets with RSA-PSS authenticated API integration and four concurrent signal strategies—mean reversion, momentum, market-making, and rule-based—feeding an adaptive engine that reweights strategies from closed-trade history. Diagnosed and fixed a SQLite write-lock race condition silently dropping trade records while live orders still executed, redesigning the write path around WAL journaling, retry-with-backoff, and record-before-trade ordering, plus a reconciliation tool cross-referencing local state against live exchange fills. Runs as a persistent background service on a self-hosted Ubuntu server, polling markets every 30 seconds across economic, political, and sports categories.
PythonTrading SystemsSQLite
Jun 2026 – Present
Retail Product Shipment Weight Prediction Model
Python library predicting actual packed shipment weight on 66,224 real warehouse shipments for cartonization software—0.61 lb MAE, 92% within-1-lb accuracy, and 54% exact-match accuracy against carrier billing weight, enabling shipping labels to be pre-printed before packing. Diagnosed a billing-comparison asymmetry in the evaluation methodology that was masking true model performance, lifting exact-match accuracy by 23 points without retraining; benchmarked Ridge against HistGBT, LightGBM, XGBoost, and CatBoost, and fixed a critical train/serve feature-skew bug.
Built a Django plugin and Flask daemon automating secure lab data access at Princeton's Micro-Nano Fabrication Center, replacing USB workflows with SMB drive mapping via HMAC-signed API communication and SQLite session persistence. Wrote a backfill script provisioning 376 existing MNFC users from NEMO API exports.
DjangoFlaskSystems
Dec 2025 – Jan 2026
AI-Driven News Headline Analysis
NLP pipeline analyzing 12,000 headlines across 50 news sources to study how AI coverage frames human behavior—embeddings, clustering, and sentiment analysis via VADER, TF-IDF, and PCA. Used Qwen2.5-7B-Instruct to generate interpretable cluster labels and insights. National finalist at RAISE 2026.
NLPData SciencePython
Dec 2024 – Mar 2026
VR Digital Twin
Co-led a 6-person cross-institutional team building a VR digital twin of Princeton's Advanced Device Packaging Lab using photogrammetry, Blender, and Unity with physics-accurate mechanics. Deployed on VR headsets via HyperSkill with three integrated AI agents; co-authored a peer-reviewed publication in the Journal of Advanced Technological Education.
Full-stack RAG-powered research assistant: upload PDFs and chat with papers grounded in actual content, with auto-generated summaries, keywords, and methodology extraction. Cross-paper semantic search via Qdrant Cloud vector DB and Hugging Face embeddings, with Firebase token auth on the backend for secure multi-user access.