Helps transplant patients understand their personal risk of organ rejection based on their medications and history
Python 3.12
Google ADK
Gemini 2.0 Flash
Cloud Run
Firestore
Problem: Transplant patients juggle complex medication schedules where one missed dose or bad interaction can trigger rejection, and they have no easy way to see their personal risk.
Solution: A team of five AI assistants work together to check drug interactions, score adherence, and estimate rejection risk, grounded in real transplant outcome data so patients and caregivers get answers they can trust.
Replaced weeks of manual data entry by automatically gathering business info for 11,000+ retail locations
Python 3.11+
Firecrawl
BeautifulSoup4
Pandas
Problem: A team was spending weeks hand-copying business details for thousands of retail locations from websites that actively block automation. Slow, error-prone, and soul-crushing work.
Solution: Built a resilient pipeline that handles tricky websites, recovers from failures, and cleans up messy data, delivering reliable records without anyone touching a spreadsheet.
Source Available on Request
Finds the jobs worth applying to, automatically, so you stop drowning in postings that don't fit
Python 3.12
Gemini 2.5 Flash
SQLAlchemy
Firecrawl
Playwright
Problem: Job searching is exhausting. Scanning hundreds of postings a day, most of them wrong fits, while the good ones quietly disappear.
Solution: Monitors 2,200+ companies, scores every posting out of 100 based on what actually matters to you, spots LinkedIn connections who could help, and delivers a ranked daily digest, so you spend your energy on applications, not searching.
Source Available on Request
Gives Claude a long-term memory so it remembers past conversations and picks up where you left off
<0.5ms
Search Speed (FTS5)
Python 3.11+
MCP Protocol
SQLite
FastMCP
Problem: Every new Claude conversation starts from zero. Past context, decisions, and preferences are lost, so you end up re-explaining yourself constantly.
Solution: Stores and instantly searches hundreds of past conversations locally on your machine, automatically organized by topic. Imports history from five AI platforms so nothing you've discussed with Claude is ever lost.
Lets you safely give Claude access to your Google Calendar, Gmail, and Docs, on your terms
Python 3.11+
MCP Protocol
OAuth2
Google APIs
Problem: Most AI tools that connect to your Google account are black boxes. You can't see or control what they're actually allowed to touch.
Solution: You choose exactly which Google services Claude can use (Calendar, Gmail, or Docs) and nothing else. Built with the tightest possible permissions so your data stays yours.
Ask about any meal in plain English and get instant nutrition breakdowns
Vertex AI Agent Builder
Google Cloud Functions
USDA FoodData Central API
GCP
Problem: Tracking what you eat usually means hunting through nutrition databases entry by entry, so most people give up before they learn anything useful.
Solution: Just describe what you ate in a sentence. The assistant pulls from 500,000+ foods across US and Canadian databases and comes back with personalized nutrition insights in under a tenth of a second.
Builds fair, balanced lineups for youth sports teams in seconds instead of hours
Python 3.8+
Flask
TeamSnap API
Playwright
Problem: Volunteer coaches were spending hours every week juggling spreadsheets to make sure every kid got fair playing time across positions, and still fielding complaints.
Solution: Pulls your roster from TeamSnap and automatically builds lineups that honor sport-specific rules (pitcher limits, goalkeeper rotations, equal playing time) across baseball, soccer, and hockey.
Turn your commute into productive AI brainstorming. Hands-free, voice-controlled, and safe
Android/Kotlin
Claude API
Android Auto
Problem: Long commutes interrupt creative flow. Drivers can't interact with AI assistants safely, losing ideas and project context between desktop sessions.
Solution: The only Claude app built for Android Auto. Talk to Claude while you drive, have it queue your ideas even without signal, and pick up the same project later on your laptop or desktop.
Source Available on Request
Daily outfit recommendations from your actual wardrobe. Weather-aware, occasion-aware, with wear tracking
React Native (Expo)
FastAPI
PostgreSQL
Cloudinary
Problem: Deciding what to wear eats up morning energy. People default to the same three outfits, get caught out by the weather, and never really know what looks good.
Solution: Suggests three outfits every morning, from clothes you actually own, based on the weather, your schedule, what's clean, and what you wore recently. Plus weekly style feedback from a virtual stylist.
Source Available on Request
A one-screen view of every code project's health. No more clicking between GitHub tabs
Color-Coded
Health Status
Python 3.12
Textual
GitHub CLI
SonarCloud API
Problem: Developers juggling a dozen projects constantly switch between tabs to check what's broken, what's open, and what needs attention, and still miss things.
Solution: A single terminal dashboard showing every project's status at a glance: health indicators, open issues, and code quality in one place, with one-keystroke navigation to whatever needs fixing.
Source Available on Request
DIY projection-based pool training system with desktop simulator and visual drill editor
Python 3.13
Flask
OpenCV
NumPy
Pygame
Problem: Serious pool players who want interactive practice tools hit a wall: commercial systems cost thousands and won't let you design your own drills.
Solution: A DIY system that projects drills directly onto your table, plus a desktop simulator to practice without setup. Design your own drills or use the built-in library. Works on 7, 8, or 9-foot tables.
Source Available on Request
Plans family meals around your actual schedule and tells you exactly what to buy, split by store
Multi-Store
Grocery Lists
Python 3.13
Web Scraping
Google Calendar API
Problem: Between kids' sports, work, and evening events, figuring out what to cook, let alone buy, becomes the weekly headache no one wants to own.
Solution: Pulls recipes from wherever you save them, merges the ingredients into one shopping list organized by store, and drops the meal plan straight onto your calendar around games and practices.
Source Available on Request
Age-appropriate training programs for young athletes across baseball, volleyball, and hockey
49
Sport-Specific Exercises
React 18
TypeScript
Node.js
MongoDB
Problem: Young athletes between 6 and 18 need training that matches their sport and skill level. Generic workouts bore them and miss the mark.
Solution: 49 age-appropriate exercises across three sports, delivered as two-week rotating programs. Coaches can import rosters from TeamSnap and print branded booklets with their team's logo.
Source Available on Request
Can AI implement a McGill University programming assignment from Rob Sabourin's Software Engineering in Practice (ECSE 428) class using TDD and minimal guidance
Test-Driven Development
Complex Number Mathematics
Problem: I wanted to see how disciplined test-driven development holds up when you let AI help write the code. Would the tests still drive the design, or just rubber-stamp it?
Solution: Built a working Reverse Polish Notation calculator (including complex numbers) by writing every test first, then letting the AI implement just enough to pass. Proof that test-first discipline keeps AI-assisted code honest.
Source Available on Request
Predicts whether you'll blow through your car lease mileage, early enough to actually do something about it
920km
Auto-Detected Baseline
Python
Prophet
LSTM
XGBoost
Streamlit
Problem: A real-world problem: my VW Atlas lease was tracking 7,000km over limit, a $1,275 to $9,115 penalty. Spreadsheets couldn't see the seasonal driving patterns that would determine the final number.
Solution: Three forecasting models working together to detect seasonal trip patterns and flag unusual driving, delivered in an interactive dashboard for running 'what if' scenarios and planning around the lease end.
Source Available on Request
Helps developers catch the edge cases they'd otherwise miss when writing tests
Test-Driven Development
Claude Code
Cursor
Problem: Developers writing tests only check the cases they thought of, and the bugs live in the ones they didn't.
Solution: Brings decades of proven testing heuristics into your AI coding workflow, so Claude or Cursor can suggest the edge cases, boundaries, and weird states you might have skipped, in any language.
Source Available on Request
Bridges analog mind maps and digital calendars. OCR handwritten notes into auto-scheduled events
Python
Google Cloud Vision API
React (PWA)
Problem: People who think best with pen and paper lose momentum the moment they have to re-type everything into a calendar. The planning style they love fights the tools they need.
Solution: Snap a photo of your handwritten mind map, and it reads the tasks, prioritizes them, lets you fix anything the OCR got wrong, and schedules them straight to your calendar.
Source Available on Request