Projects

Production systems, ML pipelines, and performance-focused work.

Somerville Museum Inventory System logo

Architected and shipped a production inventory management platform for the Somerville Museum.

Next.jsTypeScriptPostgreSQLServerlessVercel
  • ·Led 12 engineers through sprint cycles and code review standards
  • ·Designed architecture and relational schema
  • ·Built backend routes/serverless handlers
  • ·Introduced caching and query bundling for performance
  • ·Modernized hosting/database infrastructure
44%reduction in load times
95%reduction in annual operating costs
60%reduction in staff onboarding time
PingPoint logo

Full-Stack / Cybersecurity / JumboHack 2026

PingPoint

Real-time Wi-Fi security dashboard built at JumboHack 2026 that turns packet captures and port scans into plain-language risk insights for non-technical users.

Next.jsFastAPIRaspberry PiNmap
  • ·Built dashboard interaction flows and scan-state UX (initial scan + re-scan behavior)
  • ·Implemented Next.js API proxy routes to Pi endpoints for scan, Nmap, and LLM workflows
  • ·Wired streaming scan + analysis data into frontend state and device detail views
  • ·Configured Scapy scans and returns via FastAPI on Raspberry Pi 5
1st placeJumboHack 2026 (Cybersecurity track)

Built a production full-stack job board with role-based experiences for member vs business users.

Next.jsTypeScriptClerk AuthPostgres (Neon)Vercel
  • ·Full-stack application architecture and implementation
  • ·Clerk auth integration and role-based flows
  • ·Postgres schema design and deployment pipeline
80%increase in member engagement
50%decrease in admin overhead
JumboMap logo

Full-Stack Product Engineering

JumboMap

Interactive event/tabling mapping platform for campus operations.

ReactTypeScriptTailwindMySQLNext.js API
  • ·Built map interaction + user workflows
  • ·Implemented REST APIs for auth, verification, and DB access
~150concurrent users at Tufts Community Day
Adoptedby Tufts OCL
Reddit Politics Sentiment Dashboard logo

Built a dashboard that scrapes political forums and classifies sentiment + political leaning.

MongoDBGPT-4 APIPythonJavaScript
  • ·Multi-source data scraping pipeline
  • ·MongoDB schema + data flow design
  • ·GPT-4 integration + custom classification model
  • ·Frontend dashboard for model outputs
84.3%leaning classification accuracy
92.3%sentiment accuracy
Twitch Language Detection logo

Predicted language communities from graph structure using embeddings + supervised learning.

PythonNode2VecXGBoostNetworkX
  • ·Generated Node2Vec embeddings
  • ·Trained and tuned XGBoost classifier
  • ·Built evaluation and visualization pipeline
90.1%accuracy on 168k-node Twitch SNAP graph
0.88macro F1 across six language families

Alpha-beta pruning, move ordering, and board representation optimization.

C++
  • ·Implemented alpha-beta pruning + move ordering
  • ·Improved board representation and benchmarking
70%win rate vs initial implementation

Custom binary encoding/decoding pipeline with optimized memory handling.

C
  • ·Built custom binary encoding/decoding pipeline
  • ·Optimized memory handling
Compressed to 33% of original size
~3%quality loss with meaningful compression