Projects and tech stack

Trading strategy backtest chart shown in a code editor

How I pick a technology

I go with the simplest tool that solves the problem and that the business will still be using ten years from now. About twenty years ago I automated a Word template for a building inspection firm: a report that took five hours came down to thirty minutes, and the inspector still uses it today. No outside platform, no subscription, and the template belongs to him. That kind of longevity is what I'm after, with or without AI.

1. AI agents and automation

  • Language models (Claude, ChatGPT, Gemini) wired into business processes: email triage, invoice data extraction, report preparation. A person signs off on the result at every critical step.
  • Specialized assistants that answer from a company's own documents (RAG), for example on Quebec's RBQ contractor licensing rules.
  • GPU-accelerated local AI (CUDA, NVIDIA RTX cards) to analyze data and transcribe audio without sending it to an outside service.

2. Data and documents

  • Real estate data extraction robots (assessment rolls, land registries), stored in PostgreSQL and TimescaleDB.
  • Automated Word and Excel report generation, deployed across several franchisees with turnaround times under 24 hours.
  • Static websites (Publii, Cloudflare Workers and R2) structured so AI answer engines can cite them.

3. Industrial and real-time systems

  • 2D and 3D machine vision for defect detection in lumber, in C++ and C#, in plants running day and night (Autolog, 2005 to 2019).
  • Vision integrated into robotic assembly and test cells (Systemex Automation).
  • Optimization of routing software for postal services: large databases and overnight batch processing (GIRO).
  • Home automation and telemetry sensors connected to Oracle databases for real-time monitoring.

4. Personal projects

  • Python trading bots on the Interactive Brokers API, with hard-coded risk limits and automatic recovery after a network outage.
  • Backtesting on historical data with Pandas and NumPy, using vectorized operations.
  • Deployment in Docker containers and automated tests with GitHub Actions.

For an automation project in your business: Groupe Ordinata.