Agent systems & automation
Tool-using agents, browser workflows, MCP/API integration, human authority, long-running work, recovery, and failure handling.
We work on bounded AI, software, cloud, data, automation, and security problems where the valuable output is a changed system and an observable result.
Tool-using agents, browser workflows, MCP/API integration, human authority, long-running work, recovery, and failure handling.
Acceptance conditions, provider readback, external effects, cost/outcome attribution, and observability that answers whether the workflow actually worked.
Retrieval, enterprise knowledge, structured evidence, provenance, long-horizon context, and compression that does not turn every task into an ever-growing prompt.
APIs, backends, deployments, CI/CD, AWS, data flows, migrations, reliability, and automation without unnecessary re-platforming.
Use the $250 triage when the mechanism is unclear. Use the rescue when the implementation boundary is already credible.