Experience
Work history and projects, and how they connect to my main skill areas. Scroll to see which entry's skills are highlighted, or click one to pin it until you scroll again; hover a skill to see everywhere it shows up.
Web Development Lab Assistant
Work- Under contract with NASA JPL, built a mission planning web app for exploring interplanetary trajectories.
- Let mission planners filter millions of trajectories interactively by optimizing the database and graphs, migrating from SQLite to DuckDB over hive-partitioned Parquet with client-side OPFS caching.
- Led development of an interactive, 3D Mars lander simulation with Three.js.
- Built a robust CI/CD pipeline for the lab with Docker, unit testing, linting, and GitHub Actions.
Ouroboros
Project- Created the core algorithms for a tool that isolates and segments nerves and vessels in terabyte-scale 3D scans.
- Shrank the data researchers handle by over 99%, processing a 2.8 TB scan in 6 minutes on a laptop.
- Ships as an Electron app with a React/TypeScript UI, a plugin system, and a Neuroglancer-based 3D viewer.
GPX.JS
Project- A modern, TypeScript-first library for parsing GPX files and converting them to GeoJSON, published to npm as @we-gold/gpxjs.
- A maintained successor to the unmaintained GPXParser.js, with full TypeScript support and a Vitest/Playwright browser test suite.
- Runs coverage checks and an npm publish workflow through GitHub Actions CI.
AI/ML Engineering Intern
Work- Built a platform on AWS to monitor the health of 30+ production AI models in one place, adopted by 3 pilot teams and shaped by 13 stakeholder interviews.
- Built ML-based drift detection that alerts teams when live data diverges from what a model was trained on, using a Random Forest classifier two-sample test.
- Tracked how data subgroups shift over time to help pinpoint the cause, using Gaussian mixture models, BIC model selection, and cost matching.
- Created a dashboard giving teams a single view of model performance, data drift, and benchmark results, built with React, visx, and FastAPI.
- Cut LLM evaluation costs by over 60% at equal confidence by tailoring questions to each model's past answers.
- Developed an environment-agnostic Python SDK for scalable data ingestion and model analysis.
ankirai
Project- A CLI published to PyPI that transforms notes (PDFs and more) into Anki flashcards quickly using any LLM provider.
- Includes a human-in-the-loop review mode to approve, edit, or reject cards before export.
#TeamGraduate AI
Project- Built AI tools that turn teachers' lessons into flashcards and study guides aligned to Namibia's curriculum.
- Reduced lesson creation time by 87.7% with AI authoring tools built into the existing website.
- Built a custom WordPress plugin, backed by a Python service (FastAPI + Pydantic AI) on Google Cloud Run.
- Implemented a RAG system stored in WordPress's own database tables, populated and queried by the plugin.
WPI Major Qualifying Project (MQP), advised by F. Murai & O. Mangoubi · Aug 2025 – May 2026
- Led development of two AI models, Scale-MGD and SparserDiff, that generate realistic networks at scales existing methods can't handle.
- Trained on networks where all prior methods ran out of memory, and made an existing model's generation step up to 116× faster with 89× less memory.
- Built RedditWalk, a new benchmark dataset, and targeted molecule and protein generation for drug discovery.
Visualize How a Guitar Works
Project- An interactive D3.js, Web Audio API, and Canvas visualization of how a guitar produces sound.
- Includes a note editor to compose a piece and hear it played back on the guitar.
ACL Injury Risk Prediction
Project- 5-person team project predicting ACL injury risk for college athletes from a Kaggle athlete injury/performance dataset.
- Addressed extreme class imbalance with SMOTE and random oversampling, and used t-SNE to visualize class separability.
WPI Greenboard
Project- 4-person team project building an unofficial carbon emissions leaderboard for WPI, estimating impact from packages delivered to the campus mailroom.
- Built a Streamlit UI backed by a FastAPI service and a PostgreSQL database, with a Dockerized deployment.
AI Engineering Intern
Work- Saved 400+ hours of manual work annually by automating surgical robot log analysis tasks with AI.
- Built an AI assistant that analyzes HUGO surgical robot logs at an expert level in minutes instead of tens of hours.
- Connected the assistant to internal data sources with MCP and hybrid RAG to answer multi-step questions.
- Coordinated numerous AI agents and data sources with LangChain/LangGraph to perform complex tasks.
- Worked in an Agile environment (Scrum and Kanban via Jira) to manage timelines and deliverables.
WPI DS 501 · Mar 2025 – Apr 2025
- Found underserved NYC neighborhoods, using distance to the nearest subway station (from GTFS data) as a proxy for accessibility.
- Forecast revenue for potential new stations with a Random Forest Regressor (R² = 0.675) over 6 years of data.
- Pinpointed potential stops predicted to earn over $4 million annually, using interactive opportunity maps.
Gittlen Cancer Research Labs · May 2024 – Aug 2024
- Built a desktop app that lets researchers analyze large 3D medical scans on a laptop by extracting key structures, reducing file sizes by up to 693× (from 132.08 GB to 190.44 MB).
- Designed a Docker-based plugin system so other labs can add their own segmentation methods to the tool.
- Achieved approx. 40× faster computation with multithreading and multiprocessing.
- Built a React/TypeScript UI with Three.js and Neuroglancer for 3D visualization of segmented scans.
Software Development Intern
Work- Developed a key-point scanning tool for micro-CT zebrafish scans using GPU kernels for improved performance.
- Designed functional tests through a collaboration with a cybersecurity professional.
- Presented the application at a lab meeting alongside novel segmentation techniques.
Freelance Web Developer
Work- Built data visualization widgets and teaching tools for clients using technologies such as p5.js and MathJax.
- Handled project planning and negotiation with 15 clients to align with their specific requirements.
- Developed maintainable codebases while adapting to varying project scopes.
Skill Areas
AI & ML
Data Visualization
Systems & Performance
Backend & Cloud
Web & Frontend
Practices & Collaboration