Agent Experience (AX) 路 Agentic Systems 路 AI Product Engineering
I build AI applications and tools that agents can use. My focus is helping them find the right tool, finish a task, and recover when something fails.
Explore Agent Experience 路 Public repositories 路 Connect on LinkedIn
| Work | What you can explore |
|---|---|
| Agent Experience | My practical field guide to agent-facing products, with a working read-only MCP service. |
| Agent-Readiness Rubric | A published v0.1 draft for reviewing discovery, action safety, recovery, and human control. |
| Awesome Agent Experience | Curated public resources on tool use, discovery, evaluation, and agent-facing interfaces. |
| Qwen on a consumer GPU | Local-inference scripts, an optimization guide, and recorded results for a specific hardware/model setup. |
- Agent-facing tools: MCP, APIs, CLIs, reusable skills, onboarding, and actionable errors.
- Evaluation loops: realistic tasks, trace evidence, failure analysis, and fix-and-rerun workflows.
- Agentic systems: stateful orchestration, sandboxed execution, observability, and artifact verification.
My earlier work in ML infrastructure, search, and computer vision grounds these systems in reliability, latency, cost, and product constraints.
How I work: check the completed task, not just whether the answer sounds right. Make access, failures, and the points that need a person clear.
Use the field guide from an agent
The public, read-only MCP endpoint is https://agentexperience.tech/api/mcp.
It exposes search_guides, get_guide, get_readiness_rubric, and get_glossary_term. See the discovery manifest for the published interface.
Some professional work is not public. I share reusable engineering patterns without disclosing client identities, private code, or internal results.