Powell River, BC · Available for projects
Building the operating system for AI agents.
Army mechanic turned self-taught developer. I build open source infrastructure that makes AI agents smarter, cheaper, and more reliable — memory systems, context management, predictive execution, and multi-agent orchestration. One interconnected ecosystem, all MIT-licensed.
Four open source projects that work together. Each solves a distinct problem. Together, they form an intelligence layer for AI agent systems.
What agents did. Persistent memory with semantic search, dream cycle consolidation, and knowledge graphs.
NestJS · PostgreSQL · pgvector · 2,816 testsWhat agents can do. Context management that loads the right capabilities at the right resolution. 98% token reduction.
TypeScript · 105 tests · 100% recall · 64.7% precisionWhat agents will do. Bayesian prediction + multi-armed bandits for model routing, constraint injection, and cost optimization.
TypeScript · 56 tests · Grade A benchmark · 81.5% prediction accuracyThe orchestration layer. Multi-agent pipelines with approval workflows, client profiles, and sector-aware agents.
Mastra · Next.js · 92 agents · 143 ACR capabilitiesEngram remembers → ACR loads → AWM predicts → Forge executes
Something new is being built on top of this stack. Stay tuned.
One optimization loop. Three systems. Every one improved. The method is stupid simple and works on anything you can measure.
A devlog about building Engram — and the two AI agents who helped build it. Three wipes, 70 tickets, and the line between tool and teammate.
I asked my AI what we did today. It had no idea. So we taught it about time — and in the process, discovered why memory is harder than anyone thinks.
Open source memory infrastructure for AI agents. Ensemble vector search, Dream Cycle consolidation, knowledge graphs — self-hosted or cloud.
Open source memory infrastructure for AI agents. Self-hosted or cloud. Ensemble vector search, nightly Dream Cycle consolidation, knowledge graphs, and a monitoring dashboard. 8,200+ memories in production, 2,816 tests. The memory layer of the ecosystem.
Agent Capability Runtime — the missing context layer for AI agents. Manages what goes into an agent's context window: which capabilities are loaded, at what resolution, within what budget. 98% token reduction at cold start, 100% recall, 64.7% precision. 105 tests.
Agent Workflow Model — predictive execution for agent pipelines. Uses Bayesian statistics and multi-armed bandits to predict outcomes, route to cheaper models, and pre-inject constraints that prevent failures. Grade A benchmark, 81.5% prediction accuracy, self-optimizing via autoresearch.
Monte Carlo simulation engine for sports playoff brackets. 50,000 simulations, EV-optimized picks for your pool's scoring format. Live for March Madness and NHL playoffs. Data-verified with 263 automated integrity checks — every stat cited, every claim verifiable.
Lightweight CRM built for nonprofits who want real donor relationships, not automation funnels. Guided touchpoints prompt staff to reach out at the right moments — birthdays, anniversaries, follow-ups. I own the full stack: architecture, infrastructure, and shipping features. Anti-AI-slop by design.
Sales intelligence platform for healthcare. AI-powered research tools that help medical device sales teams find the right prospects faster. I work across the landing site, web app, and own the Salesforce integration.
Hired into a new role mid-Salesforce migration. Ended up owning Salesforce admin, dev, and automation — plus Shopify, Pardot, and FormAssembly. Worked closely with donor relations to improve retention: 40% fewer lost monthly donors over 2 years. Built a formal data management plan that still guides the org today.
Helped rebrand Vancouver Running Company into a global run membership powered by NFTs. Coordinated brand activations at On's NYC flagship during NFT.NYC and a Vancouver event with custom RFID-tagged shoes that triggered NFT animations on scan. Also shipped frontend work.
Gamified heart health app with pixel-art style and deep progression systems. Started as PM, grew into leading dev and product. Cut deployment time from weeks to under 10 minutes via TestFlight by building out proper test suites and CI automation. Shipped to the App Store. Project sunset in 2025.
I got my first computer at 5. By 14, I was studying IT at a technical college. By 17, I had a university offer before I'd finished high school.
I dropped out. Too much theory, not enough building.
So I built things anyway — just not software. I fixed vehicles in the Australian Army. I repped the fastest-growing running brand in Western Canada. I worked a dozen jobs that taught me how to learn anything, fast.
When my wife got pregnant with our first, I bet on myself. Turns out the kid who kept breaking his dad's computer could actually ship code for a living.
I don't specialize. I solve problems. Whether that's React, Salesforce, AI tools, or something I'll learn next week.
I help teams build agent memory systems, multi-agent pipelines, and AI infrastructure that ships to production. From architecture to deployment.
Get in touch →Got something to build? I'm probably available.