Open-source dependency-upgrade checker for developers and coding agents
- Built a CLI and GitHub Action that downloads both versions of an upgraded dependency and finds the affected call sites by file and line, across 17 package ecosystems
- Scored 83.9% precision and 97.6% recall on 16,333 real-world npm API changes from a public research corpus
- Serves its analysis to coding agents over MCP and checks AI-generated fixes file by file, reverting any that weaken tests
- Published on npm as @usedrift/cli (2,000+ downloads in September 2026), with a VS Code extension and GitHub Action
TypeScript, Node.js, MCP, VS Code API
Scout
Autumn 2026Local-first AI job-search system
- Built a local daemon, a Chrome/Safari (MV3) extension that fills application forms deterministically and stops before submission, and a React dashboard
- Exposed it to Claude through an MCP server: job search over 760k+ postings, fit scoring, deduplication and application tracking, with schema-checked tool calls and approval before anything is sent
- Generated tailored CVs on a LaTeX template, constrained to verified experience
TypeScript, React, SQLite, MCP, Chrome Extensions
SQL Interpreter & Query Optimizer
Winter 2024SQL interpreter and query optimizer to improve query efficiency
- Developed an SQL interpreter and query optimizer in Java to improve query efficiency and data retrieval
- Applied advanced data structures and optimization techniques to enhance performance
Java, Data Structures, Query Optimization
Holoquiz
Winter 2024Health-oriented AI powered quiz making game for students
- Wrote all the software in a 6-person team: Raspberry Pi button and touchscreen handling, and a Python pipeline that generates quiz questions with GPT as structured JSON and grades answers live
OpenAI API, Python, Raspberry Pi
scryblr.com
Summer 2023Full-stack app for content creation using LLMs
- Developed a full-stack app, integrating authentication, a database and an LLM API to improve UX
Flask, HTML, CSS, Javascript
PairIt
Summer 2022iOS language learning game using spaced repetition
- Built an iOS language learning game to match words across languages using SwiftUI and a custom spaced repetition algorithm
- Scraped and processed 50+ GB of multilingual data from Project Gutenberg to generate frequency-based word pairs
- Used parallel concurrency in Python to accelerate data processing by 100x, enabling scalable smart recommendations
SwiftUI, Xcode, Python
Asset Allocation
Autumn 2021Mean variance portfolio optimization
- Implemented Monte Carlo method for efficient frontier
- Modern Portfolio Theory optimization
Python, Statistics, data analysis
