AI Development That Ships
From Prototype to Production in Half the Time

Specialized in building AI applications that solve real problems and go live. 8 production applications built with proven frameworks, performance benchmarks, and maintainable code.

Results That Matter

24+ Unique Technologies
6+ 3rd Party Integrations
4 Cloud Platforms
2 MCP Servers Built

Production Projects

PRODUCTION

CRM Data Enrichment System

Production web scraping pipeline enriching 11,000+ retail locations automatically

11,479
Records Processed
10+
Fields Extracted
15+
Pollution Patterns
Python 3.11+ Firecrawl BeautifulSoup4 Pandas
Problem: Manual CRM data enrichment for thousands of retail store locations was time-consuming and error-prone, requiring extraction of structured business data from websites with anti-bot protection.
Solution: Anti-bot bypass system with Playwright-stealth, multi-strategy extraction with 6+ fallback patterns per field, checkpoint-based recovery every 50 records, and regex-based text cleaning filtering 15+ UI pollution patterns.
Source Available on Request
PRODUCTION

Claude Memory MCP Server

Searchable local storage for Claude conversation history

0.05s
Search Speed
80%+
Relevance
40MB
RAM Usage
Python 3.11+ MCP Protocol SQLite FastMCP
Problem: Claude lacked long-term conversation memory for context retrieval
Solution: Searchable local storage with full-text search, topic extraction, and weekly summaries (159 conversations tested, 7.8MB dataset)
PRODUCTION

Sports Lineup Manager

Automated lineup generation for youth sports teams with TeamSnap integration

94%+
Test Coverage
3
CI/CD Stages
SonarQube
Quality Gate
Python 3.8+ Flask TeamSnap API Playwright
Problem: Youth sports coaches spent hours manually creating fair, balanced lineups ensuring equal playing time and position rotation across multiple sports
Solution: Multi-sport platform with JSON-based sport configuration system, automated lineup generation via TeamSnap API, sport-specific rotation rules (pitcher limits, goalkeeper requirements), and 3-stage CI/CD pipeline with SonarQube quality gates
PRODUCTION

Google Workspace MCP Server

AI-native Google Workspace management

Calendar
Integration
Gmail
Integration
Docs
Integration
Python 3.11+ MCP Protocol OAuth2 Google APIs
Problem: No AI-native way to manage Google Workspace services (Calendar, Gmail, Docs)
Solution: MCP server with smart date handling and selective service enablement for minimal permissions
IN DEVELOPMENT

Multi-Sport Training Platform

Personalized workout programs for young athletes across Baseball, Volleyball, and Hockey

49
Sport-Specific Exercises
3
Sports Supported
TeamSnap
Integration
React 18 TypeScript Node.js MongoDB
Problem: Youth athletes (ages 6-18) needed personalized, engaging training programs tailored to their specific sport and skill level
Solution: Full-stack platform with centralized sport configuration system, 49 age-appropriate exercises, 2-week rotating programs, 6-page print booklets with sport-specific branding, and TeamSnap roster import
Source Available on Request
PRODUCTION

Virtual Dietitian

AI-powered nutrition analysis agent built with Google Cloud Agent Builder

500,000+
Foods Database
< 100ms
Response Time
Vertex AI Agent Builder Google Cloud Functions USDA FoodData Central API GCP
Problem: Tracking nutritional intake requires time-consuming manual database searches, preventing users from making informed dietary decisions and maintaining healthy eating habits
Solution: Conversational AI agent that instantly analyzes meals through natural language, accessing 500K+ foods from USDA and Canadian databases, delivering personalized nutrition insights in under 100ms
PRODUCTION

Polish Notation Calculator

Can AI implement a McGill University programming assignment from Rob Sabourin's Software Engineering in Practice (ECSE 428) class using TDD and minimal guidance

69%
Test Coverage
20
Unit Tests
20+
TDD Cycles
Test-Driven Development Complex Number Mathematics
Problem: Implement a Unix-style Desktop calculator using strict TDD
Solution: Successfully demonstrated AI-assisted development using strict TDD workflow, achieving functional RPN calculator with complex number support, comprehensive test coverage, and proper error handling through iterative Red-Green-Refactor cycles
Source Available on Request
IDEATION

LeaseWise ML Predictor

Machine learning system for vehicle lease mileage prediction and cost optimization

85%+
Prediction Accuracy
$1,275
Savings Projected
920km
Auto-Detected Baseline
Python Prophet LSTM XGBoost Streamlit
Problem: Volkswagen Atlas lease projected 7,084km overage ($1,275-$9,115 penalty). Traditional spreadsheet analysis couldn't account for seasonal trip patterns.
Solution: Multi-model ML ensemble (Prophet, LSTM, XGBoost) with automated pattern detection, anomaly identification, and interactive Streamlit dashboard for scenario planning and cost optimization.
Source Available on Request

Technology Stack Overview

Production-proven expertise across AI development, full-stack engineering, and cloud infrastructure

103+
Total Tech
45+
Python
40+
JS/TS
4
Databases
4
Cloud
10+
APIs
5
QA Tools

Featured Technology Stacks

Backend

Flask • Express • SQLAlchemy
PostgreSQL • MongoDB • Kafka

Frontend

React 19 • Vite • TypeScript
Tailwind CSS • Electron

AI/ML

Prophet • LSTM • XGBoost
Vertex AI • pandas

Data Engineering

pandas • Playwright • Firecrawl
Apache Kafka • PostgreSQL

Cloud & DevOps

GCP • AWS • Netlify • Render
GitHub Actions • SonarCloud

Testing

pytest • Jest • Vitest
Playwright • SonarQube

Python Development
Flask pandas pytest SQLAlchemy Pydantic black mypy beautifulsoup4 + 37 more
AI & Machine Learning
Prophet LSTM XGBoost Vertex AI
JavaScript/TypeScript
React 19 Vite TypeScript Express Tailwind Jest Vitest Electron + 32 more
Databases
PostgreSQL MongoDB SQLite Kafka
Cloud & Infrastructure
GCP AWS Netlify Render Docker Cloud Functions
APIs & Integration
Google Workspace TeamSnap USDA API MCP OAuth2 + 5 more