← Back to portfolio

Currently at SwingVision

Jayden Lee

Junior · UC Berkeley, Data Science

ML systems for messy, real-world data: fan behavior at Cal Athletics, LLM benchmarks at MatX, proteomics at a national lab.

Jayden Lee

I'm a junior studying Data Science at UC Berkeley, working at the intersection of machine learning, software, and sport.

Data Analyst Intern at SwingVision. Collaborating with ML engineers on labeling workflows for sports data.

Study
B.A. Data Science, Applied Mathematics emphasis, UC Berkeley
Degree
Expected May 2028
GPA
3.7
Languages
Python, SQL, R, Java, C++, JavaScript, TypeScript, HTML/CSS
Data & ML
Pandas, Polars, NumPy, scikit-learn, PyTorch, TensorFlow, Matplotlib, seaborn, PostgreSQL, Parquet, ETL/ELT
Tools
Snowflake, Supabase, Docker, Git, Slurm, Tableau, Jupyter/Colab, Excel, Claude Code, Codex, Jira, conda
Software
React, Vite, REST APIs, Playwright, automated testing, Git worktrees
Activities
Sports Analytics Group at Berkeley (Head of Business) · Cal eSports D2 League Coach
Mail
jeyl@berkeley.edu
Code
GitHub profile

Lea: LiDAR Enabled Assistant

An assistant for finding hardware issues, combining iPhone LiDAR scans with object detection and vision analysis.

Built a hardware inspection pipeline combining OWLv2 zero-shot object detection with Claude vision analysis to localize components in phone-captured images and assess potential defects from cropped regions.

Co-integrated iPhone LiDAR scanning and designed an inspection interface connecting interactive 3D models, annotated photographs, and AI explanations to help users locate and understand hardware issues.

Stack
Swift · ARKit · Python · FastAPI · PyTorch · Runpod
View on Devpost
View on Devpost

bearings

A full-stack Berkeley exploration platform with authenticated check-ins, XP progression, and organization leaderboards.

PostgreSQL keeps user, location, and progression data persistent. The platform brings check-ins, XP progression, and organization leaderboards into one Berkeley exploration experience.

Connected Jira tasks to coding agents through MCP, and built a quality gate spanning TypeScript compilation, component tests, accessibility, browser interactions, visual screenshots, and change-boundary validation before integrating agent-generated code.

Stack
React · TypeScript · PostgreSQL · Playwright · Jira / MCP

Foresight

A clinical intervention simulator: a simulation and decision-support system, built to compare intervention strategies over time.

Placed first out of 50 UC Berkeley teams by developing a clinical intervention simulation and decision-support system.

Trained a Random Forest model to estimate changes in patient risk over time and dynamically compare intervention strategies for high-risk patients.

Stack
Python · scikit-learn · NumPy · Tableau
View app
View app

ChessBlitz

An open-source teaching platform in use at Berkeley Chess School.

Flask and Supabase support the storage layer, OpenRouter generates position-specific hints, and a React dashboard helps teachers manage rosters and assignments.

Stack
Flask · Supabase · OpenRouter · React
View source
View source

This website

jaydenelee.com. The front page is under works; this desktop is what is inside the wall.

Maintained a 7K+ line TypeScript codebase across dozens of AI-agent development sessions using machine-readable specifications, automated quality gates, and structured release logs.

Parallelized development by routing design, implementation, and code-review tasks to specialized agents operating in isolated Git worktrees, preventing merge conflicts across concurrent workflows.

Stack
TypeScript · Vite · Git · Claude Code · CI/CD

Experience

  1. Sep 2026 — Present

    Data Analyst Intern

    SwingVision · Berkeley, CA

    Collaborating with ML engineers to develop a labeling workflow, improving annotation consistency across more than 100 hours of sports data.

  2. May — Sep 2026

    AI Engineering Intern

    CalCompute · Berkeley, CA

    Built an agentic Second Brain knowledge pipeline that transforms raw research into structured, source-grounded notes for LLM retrieval and reasoning workflows.

    Optimized LLM context across a 14-query evaluation: prompt size fell 72.5%, inference cost fell 65.3%, and mean latency fell 33.5%.

  3. Jan — May 2026

    AI Software Engineer

    MatX · Berkeley, CA · Contract

    Engineered a Kubernetes benchmarking platform that places model submissions into standardized Docker runners, enabling reproducible evaluation without rebuilding container images for each submission.

    Automated the end-to-end submission, metric collection, and leaderboard pipeline for LLM inference benchmarks.

  4. Feb — May 2026

    Data & AI Engineer

    Cal Athletics · Berkeley, CA · Contract

    Cleaned and unified more than 3 million ticket records across 200+ attributes into a Snowflake and Colab pipeline that fed the fan engagement model used across Cal's 30+ varsity sports.

    Predicted ticket-purchase propensity from behavioral data with clustering and supervised ML models, producing a dynamic fan scoring system.

  5. Jun 2024 — May 2025

    Data Science Intern

    Pacific Northwest National Laboratory · Richland, WA

    Streamlined cross-model AI benchmarking with automated Python and Parquet ETL pipelines that unify three data schemas across 300,000+ mass spectra datasets.

    Accelerated inference across 300,000 files by scaling GPU batch workflows on a Slurm HPC cluster, and implemented target-decoy statistical frameworks to model false discovery rates.

B.A. Data Science, Applied Mathematics emphasis, UC Berkeley · GPA 3.7 · expected May 2028

Sports Analytics Group at Berkeley (Head of Business) · Cal eSports D2 League Coach

Your browser won't preview PDFs inline.

Download résumé
Kind
PDF document
Size
103 KB
Pages
1
Download résumé