2009Education
Computer Analyst/Programmer
Graduated from Red River College, Winnipeg, with a diploma in computer analyst and programmer.
Profile // Systems Architecture
17 years building software that has to work.
Full stack systems architecture across enterprise and consumer products. Currently researching LLM-assisted development to separate what works from what's hype.
2010Inventory Software Company
Full stack developer
Joined a small shop in Winnipeg building inventory management software for restaurants.
2012Inventory Software Company
Platform migration
Migrated the product from PowerBuilder 11 to .NET Framework 4, both frontend and backend.
2015Inventory Software Company
Enterprise greenfield
Built the enterprise version from scratch, connecting multiple restaurant locations to a head office. ASP.NET MVC on .NET Framework 4, hosted in Azure.
2018Inventory Software Company
COTS to SaaS migration
Migrated the product from on-premise packaged software to a SaaS model. Completed successfully and the company was acquired in 2020.
2021Agritech Software Company
Systems architect, full stack
Moved into complex agricultural projects connecting manufacturing systems together, optimizing planning and sales workflows.
2022Personal LLM projects
First exposure to large language models
Started using ChatGPT shortly after its release. Initially treated it as an advanced chatbot and used it for creative writing and personal experiments. The output was generic, but the trajectory was clear. Learned that a model responds to its current context rather than learning from an individual conversation.
2023Agritech Software Company
Open source and cloud migration
New projects and features built on open source technologies and cloud infrastructure. Picked up GCP, Python, and various JS frameworks alongside existing .NET Framework 4.5 systems.
2023Personal LLM projects
From chat to coding assistance
Used GitHub Copilot for autocomplete and ChatGPT as a Python tutor while building personal software projects. Created a repeatable benchmark for testing whether models could produce complete, working programs rather than only convincing conversation. Tested local models with Ollama and began using aider, my first LLM coding agent. Direct API usage also made the cost of hobby experimentation clear.
2024 H1Personal LLM projects
From assistants to coding agents
After seeing Gemini at a Google conference in Las Vegas, continued testing new models with a repeatable software benchmark. Moved from aider to Cursor as personal projects became more ambitious and subscription pricing made sustained hobby experimentation more practical. Learned React, Python backends, and Next.js while most models still failed the benchmark.
2024 H2Personal LLM projects
Agentic development and capability testing
Shifted from using an LLM as an assistant to directing it through larger tasks in personal projects. Reasoning models began solving the benchmark, a meaningful capability change after repeated failures. Focused on browser-based Three.js games as a hobby while testing Replit, Bolt.new, Windsurf, and other tools. Generative LLMs improved quickly across several modalities, but the cost of sustained experimentation became difficult to justify.
2025Agritech Software Company
.NET Framework to .NET 10 migration
Migrated an internal application from .NET Framework to .NET 10 with a new Vue 3 frontend. A long-term goal I was able to see through. The application can now run in the cloud and be developed on from any platform, with future SaaS potential.
2025 H1Personal LLM projects
Testing the limits of the hype
Tested DeepSeek R1 and found its reputation ahead of its usefulness for personal projects. It still exposed a broader issue: model capability was improving while the industry’s spending expectations were becoming harder to reconcile. Reduced subscriptions and focused on personal web projects and browser-based Three.js games as a hobby. LLM-generated visual assets remained less reliable than premade assets and Mixamo for these experiments.
2025 H2Personal LLM projects
Workflow integration and model evaluation
Built browser-based Three.js games as a hobby and explored crypto projects. Claude Code changed my personal-project workflow by moving LLM assistance into the terminal instead of tying it to a VS Code-based editor. Became more deliberate about evaluating LLM-generated code and models by intelligence, context handling, cost, and sustained performance. Basic CRUD software was becoming increasingly easy to reproduce in personal projects, shifting value toward architecture, domain knowledge, and execution.
2026SaaS Startup
Livestock production systems
Integrated systems software for livestock production. Optimizing plant production planning, maximizing sales, full inventory utilization, and logistics tracking.
2026Current personal projects
Controlled agent workflows and local inference
Focused on practical agent workflows, model selection, and control over compute and data for personal projects. Prefer agents that can run from a local terminal and operate within a known project. Open models are becoming credible for personal development, while local inference offers a path for keeping project code and data under my control. Exploring what a small home lab can make possible while keeping architectural judgment with me.
About FAQ
Before you read
- Who is this site for?
- People building or evaluating software systems, especially where LLM assistance changes decision-making, architecture, or execution quality.
- What kinds of work does Ryan write about?
- LLM-assisted development, coding agents, systems architecture, software moats, tooling strategy, and practical production lessons.
- Why focus on LLM and systems architecture together?
- Because the value shift is happening in both directions: stronger tooling reduces some implementation costs, while long-term durability still comes from architecture and judgment.