Cedrick Castro
Back to work

Kapitol / Senior Software Engineer

Workflow reliability

Background jobs, credit consumption, recovery paths, typed APIs, and logging for long-running agent workflows.

01

Moved long-running generation work into explicit background jobs with clearer state and logging.

02

Built credit consumption and recovery paths around generated-message workflows.

03

Improved endpoint wiring, schema correctness, loader context typing, and worker API boundaries.

Problem

Agent workflows are long-running and failure-prone: generation, review, sending, replies, and billing-related credit use all need observable state.

Decisions

The platform used explicit jobs, worker API boundaries, typed schemas, logging, and recovery paths so generated outreach could be retried, inspected, or corrected.

Outcome

Kapitol became easier to operate and extend across lead processing, message generation, approvals, follow-ups, replies, and credit-backed usage.