Work
Case studies
Case studies by Felipe Cardoso: how AI agent systems are designed, evaluated and run in production.
-
Running a fleet of coding agents
How I plan, delegate, review and verify the work of AI coding agents, so that more parallel work does not mean less control over quality.
Method and daily practice
- Hermes Agent
- Claude
- GPT
- DeepSeek
- local Qwen models
- Git and CI
-
A local inference and evaluation lab
How I serve and benchmark open models on my own hardware, with a harness where every number in a report traces back to a stored request.
Sole builder and operator
- Python
- llama.cpp
- Strata
- Vulkan
- BATS
- systemd