Behind the work / The Atelier
Welcome to
the atelier.
A look inside my engineering workshop: the systems I build, the tools I use, and the experiments that shape how I work with AI.
$ show workbench
What’s on the whiteboard
Explore the decisions behind a selection of independent systems.
Bound the tools.
Multi-agent workflows with role-specific tool permissions, structured inputs, and output validation.
Explore project ↗02 / PLATFORMShip through Git.
A hybrid Kubernetes platform with GitOps, private networking, managed secrets, and observability.
Explore project ↗03 / PRODUCTKeep the till working.
A multi-tenant application with database isolation and an offline transaction queue.
Explore project ↗04 / FOUNDATIONMake it repeatable.
Machine-readable contracts and deterministic application scaffolding across repeated runs.
Explore project ↗$ trace a memory
From data to agent context
Explore how information can travel from a source through normalization and retrieval to an agent.
- Source
- Normalize
- Embed
- Store
- Agent
{ "type": "task", "status": "done" }
{ "entity": "task", "state": "completed" }
The agent can use a consistent record across sources.
Illustrative examples—not live personal data or a live model response.
$ inspect inference
The local inference rig
I run a self-hosted GPU inference bench for local models and embeddings, exploring quantization, model routing, and memory budgets.
Architecture overview. The inference service is not exposed on this page.
$ review boundaries
A security-minded workshop
[scope] Role-specific tool permissions
[validate] Schemas and deterministic checks
[test] Prompt injection and escalation scenarios
[observe] Logs, boundaries, and failure analysis
Illustrative design principles from the security lab; not a live security feed.
$ query the résumé
Ask the workbench
The interactive résumé uses the same MCP service available to recruiting agents.
Connect your own agent ↗Start with a good question.
Explore the same evidence your AI agent can access.
Project descriptions and experience are candidate-provided. This résumé's MCP tools are directly testable; other systems are available for an architecture walkthrough. Current and past employment positions and employers are private. Professional focus and target roles do not disclose a current job. Independent verification and references require direct contact.