I build and explore agent systems — how they reason, use tools, remember context, recover from failure, and turn intent into useful action.
This profile is my small workshop for practical experiments, unfinished ideas, and systems that get a little more capable with every iteration.
|
Planning loops, tool calls, state transitions, evaluation, and the engineering that keeps an agent moving. |
RAG, memory, retrieval, and choosing the right context instead of simply adding more context. |
Connecting agents to APIs, data, workflows, and the messy real world beyond the chat box. |
|
|
|
There are still better ways for agents to plan, use tools, remember, collaborate, verify their work, and recover when reality does not match the plan.
I want to build some of those missing pieces — patiently, practically, and one working system at a time.
Build the loop. Test the edges. Keep the useful parts.