Got my first computer at 6 — an Oric Atmos 48K. Never really stopped.
Architecture is about reducing complexity for the teams that come after — not just picking the right tools. Systems should remain understandable over time — by humans, not just machines.
Data Architect · Systems Thinker
Rust · Python · AWS · Terraform · Databricks· Snowflake
Normandy, France
I design and build systems where AI is engineered, not improvised. Three principles guide most of my work:
- Knowledge is a first-class system
- Agents are structured operators, not autonomous chaos
- AI must be engineered — not just prompted
Most of my projects map to this flow:
Knowledge → Context → Execution → Agents → Observability
The first objective is not autonomy. It is control.
-
llm-wiki — Git-backed wiki engine with MCP server — documentation that compounds over time
- llm-wiki-skills — ingestion, research, and knowledge graph workflows
- llm-wiki-hugo-cms — render your wiki as a static site
- asdf-llm-wiki — asdf plugin
-
Agent Systems (exploration — early days) — patterns and tooling for observable, controlled agent systems
- async-btree — Async behavior tree — composable, awaitable control flow
- networkx-query — Query layer for NetworkX graphs
- async-rx — Reactive programming for async Python
- terraform-aws-tf-registry — Private serverless Terraform registry on AWS
- terraform-aws-tf-registry-cli — Python client for terraform-aws-tf-registry
- petgraph-live — Generation-keyed graph cache with persistence and algorithms
- qjl-sketch — CPU-efficient vector compression with strong accuracy trade-offs
Becoming a beekeeper. First swarms incoming.
Turns out distributed systems also exist in nature.
All projects and opinions here are my own — nothing here represents or engages my employer.