Projects
Side projects where I stay sharp, test ideas, and ship working products.
AI & Automation
Sleeper Lock-In Optimizer
AI-powered fantasy basketball decision system (2026)
Built a decision-support tool that analyzes fantasy basketball lineups and recommends daily lock-in decisions. The system pulls real-time data from multiple sources - player performance trends, upcoming schedules, injury reports, matchup quality - and synthesizes them into actionable recommendations delivered via daily email.
The interesting problem: fantasy decisions involve weighing multiple uncertain factors with different time horizons. A player trending up might have a brutal schedule next week. A favorable matchup means nothing if someone's nursing an injury. The system handles this by scoring each factor independently, then combining them into a single recommendation with transparent reasoning.
Handles production realities like API rate limits and blocked IPs through caching and graceful degradation.
Python, Sleeper API, SQLite, GitHub Actions
Todoist AI Assistant
AI prioritization and routing for task management (2024)
Supercharges Todoist with intelligent prioritization, automatic labeling, and section routing. Connects LLM reasoning to existing productivity workflows without requiring behavior change.
Python, OpenAI API, Todoist API
StarFormatter
Format text for AI prompts (2024)
Live tool for preparing text for AI consumption. Strips formatting, normalizes structure, handles edge cases.
Astro, Vercel
Tools & Utilities
Fresh Tabs
Chrome extension for tab hygiene
Start every day with a clean browser. Automatically archives unpinned tabs daily or on demand.
Chrome Extension APIs, JavaScript
Product Concepts
Quillworks
Passive capture product thesis
A product thesis exploring how to capture digital value without interrupting analog behavior. Wrote design principles and scoped a hardware roadmap for a pen add-on that tracks writing without mode switching. The thesis clarified a principle I keep returning to: the best digital tools earn trust by disappearing into existing habits.
More experiments live on GitHub.