AWS Step Functions / 2025 to 2026

Operational and agent tooling

I built a command that consolidates customer-impact signals across four services. I also shipped agent workflows for team task routing, with write confirmation, duplicate handling, and 21 deterministic test scenarios.

Customer-impact reporting

Incident response required engineers to assemble customer-impact data service by service. I built a read-only command that collects seven signals across four Step Functions services and returns a regional summary with exception details.

I made incomplete or unavailable data explicit, avoided adding customer counts that could overlap, and preserved historical counts when detailed contributor lists had expired. The tooling used scoped cross-account reader roles and included regression tests for historical replay.

Shared agent workflows

I moved task-routing workflows from personal tooling into the team's agent package. The workflows classify tasks, preserve project placement and labels, handle duplicates, require confirmation before writing, and verify the result afterward.

I shipped 21 deterministic scenarios covering routing, placement, duplicate handling, opt-out behavior, and label safety. I also shipped an operating-procedure workflow for retention-decrease requests.

Tools and team ownership

My operational tooling work also includes quota batching, validation, retries, and review requirements for production capacity changes.

I help teammates adopt the tools through code review, onboarding, and runbooks. Earlier work included mentoring engineers on CloudFormation ownership and sharing region-build context with new launch owners.

Technologies

Python · IAM · Operational tooling · LLM agents · Deterministic testing