Timeline
Sports scientist to AI engineer. The full arc — research published, products shipped, and where I'm headed.
2026
AI Eval Lab deployed + 7-Tool CLI Toolkit complete
Shipped AI Eval Lab (systematic prompt evaluation: define test cases, score prompts, AI-powered fix suggestions). Completed the 7-tool CLI dev toolkit: ai-commit, smart-pr, ai-explain, ai-review, ai-changelog, project-context, ai-standup. The toolkit now covers the full dev workflow from commit to standup.
Multi-Agent Coordination Demo — Live
Five-agent pipeline made visible: Planner decomposes a task into subtasks, three Specialists (Researcher, Writer, Critic) execute in parallel streams, Synthesizer assembles the final output — all streaming in real time. Built to make multi-agent orchestration tangible, not a diagram. Portfolio now at 35 apps, 28 blog posts.
Open to opportunities
Actively looking for the right AI engineering role — somewhere building real things with real impact. Focus: Vancouver-based or remote-Canada AI-native product companies, AI ops, LLM systems, production ML.
dudleyrode.com/ask — AI Recruiter Chat
Added a live AI chat to the portfolio: recruiters can ask anything about Harrison's background, projects, papers, and stack. Claude Haiku responds with full context. Shows recruiters AI-native development in action, not just on a resume.
interview-prep AI — Live
Built interview-prep: paste any job description → get 12-15 tailored interview Q&A in 5 categories (technical, behavioral, situational, culture, strategic). Powered by Claude Opus with streaming. Harrison's own background baked in as system context.
clip-finder — Live
Built clip-finder: YouTube AI highlight detector. Drop a URL → get the 5-7 most shareable moments, with timestamps, suggested captions, and platform recommendations (TikTok vs LinkedIn vs Twitter). Built in 2 hours. Domain-relevant to OpusClip.
90-Day Build Sprint: 15+ Apps Shipped
January 2026: accelerated output phase — building and deploying a new app roughly every 3-4 days using the Matua + Claude + Codex AI-native workflow. Apps include: workout tracker, rep-counter (MediaPipe ML), finance dashboard, collab whiteboard, flight scanner, snow forecast, currency converter, AI form analyzer, and more. All live.
2025
NZ Adventure Planner — Live
Built the NZ Adventure Planner: real-time condition scoring for New Zealand hiking tracks using Open-Meteo weather data and GeoNet earthquake/volcano feeds. Competes directly with DOC's official tool on UX.
Built MCP Servers + Custom Tool-Calling Infrastructure
Designed and deployed a suite of Model Context Protocol (MCP) servers enabling Matua to call external tools, access APIs, and coordinate multi-step workflows. Set up standardized tool-calling patterns now used across all projects.
Durability Analyzer — Live
Built and launched the Durability Analyzer: a client-side Garmin FIT file analyzer based directly on own peer-reviewed research. First app to translate the decoupling + VT1 drift papers into an interactive tool. Zero data leaves the browser. The portfolio centrepiece.
Paper 2: Decoupling Prediction in Elite Endurance Athletes
Second peer-reviewed publication demonstrating that cardiac decoupling during prolonged exercise can predict sustainable race intensity with high accuracy. Built the mathematical model underlying the Durability Analyzer's efficiency scoring algorithm.
2024
Built 'Matua' — AI Co-Developer Agent
Created and deployed Matua: a persistent AI agent (Claude + Codex) that functions as a software co-developer. Matua handles code writing while Harrison handles architecture, review, and product direction. The AI-native dev workflow that made the following 90-day ship sprint possible.
Paper 1: CHO Supplementation & VT1 Shifts During Prolonged Exercise
First peer-reviewed publication on carbohydrate supplementation effects on metabolic thresholds during prolonged endurance exercise. Demonstrated that VT1 drifts meaningfully with fatigue — directly informing the durability concept later coded into the Durability Analyzer.
2023
Started at Supermix as Operations Manager
Joined Supermix to lead AI operations — managing LLM systems in production, building AI-native workflows, and shipping internal tools. The role evolved into building AI infrastructure from scratch, including the 'Matua' agent orchestration system.
Joined Athletica AI as Coach
Joined Athletica AI — an AI-powered training platform — as a coach, bridging the gap between exercise science and machine learning. First deep exposure to LLMs integrated with physiological domain logic. The pivot from pure research to AI+sport began here.
2022
Graduated: Master of Sport, Exercise & Health
Completed MSEH with a focus on endurance physiology and performance prediction. Thesis work in metabolic modeling laid the scientific groundwork for the Durability Analyzer built years later.
2021
Joined SPRINZ as Research Associate
Began working with the Sports Performance Research Institute New Zealand (SPRINZ), contributing to high-performance sport research. Led experiments on physiological modeling, endurance limits, and metabolic thresholds.
2019
Started Masters in Exercise & Sport Science
Enrolled in the Master of Sport, Exercise & Health at Auckland University of Technology (AUT). Focus: endurance physiology, performance modeling, and sensor-based measurement. The foundation that would eventually connect to AI.
The beginning
What's next is being written now.