Chanbin Park

CS @ Princeton

Education

Bachelor of Arts in Computer Science

Minors in Statistics & Machine Learning and Visual Arts

Loading contributions...

cp5721@princeton.edu

Open to opportunities

--Applied TensorFlow ML/NLP embeddings to analyze 30K+ conversational text segments and quantify social patterns.
--Built Python pipelines (pandas, NumPy) to process 50K+ embeddings reducing preprocessing time by 40%.
--Used scikit-learn to evaluate embedding quality through clustering and cosine similarity analysis of conversational text.
--Designed modular configuration-driven analysis code reused across 3+ research projects to reduce duplicated logic.

--Developed a full-stack “Guess the Player” web app using React and Phaser deployed in production for 200+ partners.
--Designed a backend API with Express and MongoDB achieving ∼45 ms leaderboard fetch latency at 10K+ records.
--Added read-through Redis caching to leaderboard endpoints to reduce database load and stabilize response latency.
--Implemented gameplay analytics (guesses, time, opt-in email) enabling 1.2K+ submissions and a 65% consent rate

--Built a Python NLP pipeline to convert 85+ research papers into structured summaries reducing review time by 50%.
--Implemented a TypeScript and React web application for exploration of 1K+ entities with dynamic filtering.
--Developed Python workflows to normalize text into JSON datasets cutting manual preprocessing by 60%.
--Indexed extracted PDF text using SQLite FTS5 to enable fast keyword search across 750+ research documents.

--Engineered an AI video surveillance platform that interprets live camera feeds and autonomously triggers actions via MCP tools winning SB Hacks XII (Grand Prize, 2nd Place) among 340+ participants.
--Combined on-device YOLOv8 detection achieving 85% precision with a from-scratch blob-tracking motion algorithm maintaining stable object tracking across 90%+ frame transitions at 1s/frame.
--Built an event-driven pipeline storing 5K+ timestamped detections in MongoDB Atlas for real-time querying.
--Designed a modular provider system with runtime switching and local fallbacks for reliable MCP tool execution.

--Developed a real-time collaborative video editor with WebGPU/WGSL supporting advanced temporal effects.
--Implemented a CRDT-based conflict-free collaboration system using Yjs and WebSockets with sub-100 ms latency.
--Engineered a Docker-containerized Go backend for FFmpeg processing and AWS S3/CloudFront delivery.
--Developed a CI/CD pipeline using GitHub Actions for automated builds, validation, and Vercel/AWS deploys.

--Built a web app that turns lectures and articles into AI notes and compiles friends’ notes and Q&A into one master doc.
--Implemented a real-time audio transcription pipeline using API routes and WebSockets that ingests audio/transcripts and uses Google Gemini for voice-to-text transcription from sources like Zoom and YouTube.
--Utilized Grok (xAI) to generate notes and power live Q&A as well as fold answers back into the running notes.
--Designed Supabase/Postgres schema with RLS and Supabase Auth deployed on Vercel with env-scoped configs.