Projects

I build tools that multiply what people can achieve—from enterprise AI systems that give mining teams instant access to operational knowledge, to privacy-first productivity tools that run entirely on local hardware.

Featured Projects

Three systems that showcase the range of capability: enterprise AI, local-first privacy, and production infrastructure.

Mining Knowledge AI Platform (OGC-MOS)

Complete AI-powered knowledge management system for mining operations. Enables AI agents to semantically search across 1300+ KPIs, operational procedures, and visual content.

Architecture

  • OGC-MOS-MCP: Production FastAPI server with 5 MCP tools deployed on Railway
  • OGC-MOS-Embeddings: 32-script pipeline for hierarchical vector embeddings (3 tiers + visuals)
  • OGC-MOS-Docs: Docusaurus documentation platform with AI synthesis agents

Technology Stack

PythonFastAPIQdrant CloudOpenAIDocusaurusRailway

Impact

Reduces time-to-answer for operational questions from hours to seconds. Preserves institutional knowledge and enables AI-augmented decision-making.

Local Meeting Transcription System

Privacy-first meeting transcription running entirely on Apple Silicon. No cloud dependencies, complete local processing.

Architecture

  • Swift menu bar app with Python core daemon communicating via Unix sockets
  • MLX Whisper large-v3 for transcription optimized for M4 Max
  • Pyannote-audio for speaker diarization with voice learning
  • Llama 3.3 70B for AI summaries and action item extraction

Technology Stack

PythonMLXSwiftLlama 70BPyannote

Impact

Full meeting transcription with speaker identification and AI summaries without sending any data to the cloud. Optimized for M4 Max with 128GB RAM.

Personal MCP Server Infrastructure

Production-grade MCP infrastructure with Auth0 OAuth, Cloudflare Tunnels, and multiple deployed servers for AI-augmented productivity.

Architecture

  • rainworth-mcp-servers: Multi-server monorepo with shared auth and deployment config
  • Mealie MCP: Meal planning assistant with recipe management
  • UP Bank MCP: Australian banking data access for financial AI
  • PowerPoint MCP: Semantic search across presentation content

Technology Stack

PythonFastAPIAuth0Cloudflare TunnelsQdrantSystemd

Impact

Demonstrates production MCP deployment patterns with authentication, monitoring, and self-hosted infrastructure. Enables Claude to access real-world data sources.

Enterprise Systems

AI systems and applications built for mining operations. Production deployments with real users.

M&A Evaluation Agent System

Multi-agent system for mining M&A due diligence built on Google's Agent Development Kit with 4 specialized sub-agents.

  • Orchestrator agent coordinating technical, financial, regulatory, and comparables analysis
  • Custom MCP tool servers for DCF modeling, commodity pricing, and transaction multiples
  • Terraform infrastructure-as-code deployment on Google Cloud
PythonVertex AI ADKGemini 3 ProTerraformGCP

OKR Tracking Application

Full-stack application for strategic objective management with real-time collaboration and drag-and-drop interface.

  • Strategy → Initiative → Key Action hierarchy with progress tracking
  • Supabase backend with Row Level Security for multi-user access
  • Visual indicators and dashboard for portfolio-level view
Next.js 14ReactTypeScriptSupabaseTailwind CSS

Tailings Governance Agent

Microsoft 365 AI agent for tailings compliance monitoring and governance automation.

  • Deployed via Microsoft Teams with SharePoint and Office 365 integration
  • Automated compliance checking and documentation workflows
  • Built with Microsoft 365 Agents Framework and Teams Toolkit
Microsoft 365 AgentsTeams ToolkitSharePoint

Personal Projects

AI automation and productivity tools. Privacy-first where it matters, cloud-enabled where it helps.

personal-os

AI-augmented life management system integrating Claude with productivity tools.

  • Google Calendar, Todoist, Obsidian, Whoop, UP Bank, Gmail integrations
  • A "cognitive prosthetic" for managing complexity without losing context
  • MCP ecosystem enabling AI to access real-world data sources
MCPClaudePythonMultiple APIs

Handwriting OCR

Local-first tool for extracting text from handwritten notes in Obsidian.

  • 100% local processing using MiniCPM-V-2_6 via LM Studio
  • Automatic image discovery and batch processing
  • Backup creation and progress tracking for vault safety
PythonMiniCPM-VLM StudioLocal AI

Telos (FreedomTracker)

Financial independence tracker with AI-powered banking integration and FIRE calculations.

  • UP Banking integration with Gemini-powered transaction analysis
  • Automatic profile pre-fill from 12 months of transaction history
  • Scenario planning and time impact analysis for financial goals
Next.js 16SupabaseGemini 3.0 ProUP API

PowerPoint MCP Server

Semantic search server enabling Claude to search and retrieve PowerPoint slide content.

  • 5 MCP tools: semantic search, full retrieval, design patterns, research, outline generation
  • Qdrant vector database with Google Gemini embeddings
  • Railway deployment with SSE transport and bearer auth
FastAPIPythonQdrantGeminiRailway

Technology Stack

Full-stack capability across backend, AI/ML, data, and infrastructure.

Languages & Frameworks

Python
FastAPI, data pipelines, AI/ML
TypeScript/JavaScript
Next.js, React, Node.js
Swift
macOS native applications

AI & ML

OpenAI
Embeddings, GPT models
Google Gemini
Vertex AI ADK, multi-agent systems
MLX
Apple Silicon optimized Whisper, Llama
Anthropic MCP
5+ production servers deployed

Data & Storage

Qdrant / ChromaDB
Vector databases for semantic search
PostgreSQL / Supabase
Relational data, auth, RLS
Snowflake
Enterprise data warehousing

Infrastructure & Deployment

Railway / Vercel
Production app hosting
Cloudflare Tunnels
Self-hosted service exposure
Terraform / GCP
Infrastructure as code
Systemd
Linux service management

Key Differentiators

  • Production MCP Expertise: 5+ deployed servers with auth, monitoring, real users
  • Privacy-First AI: Local transcription and OCR on Apple Silicon
  • Full Enterprise Systems: Not prototypes — Qdrant, Railway, Vercel production deployments
  • Mining Domain + AI Depth: Purpose-built tools requiring deep domain knowledge