AI-Augmented SDLC
Agentic tooling that drains the toil out of a small team running a large fleet - with humans keeping every approval gate.
- Role
- Sole author
- Period
- 2024 - present
- Ownership
- Sole author
- Stack
- AWS Bedrock (Claude) · GitHub · Jira · Claude Code plugins · Python · S3/JSONL
Problem
Five engineers, eight production tenants: triage, code review, and backlog grooming eat the week if you let them. The interesting question wasn't whether AI could draft a fix - it was how to wire it in so speed goes up while accountability stays exactly where it was.
Approach
Every pipeline is built around human gates. Agents propose; people approve. State lives in labels on the ticket and PR, so any step is inspectable and reversible, and nothing merges without a named human owning the decision.
Architecture
Auto-remediation: an AWS Bedrock service (Claude via bedrock-runtime) triages production alerts, correlates them with recent changes, and opens a fix PR for review.
Review: a Claude-based PR reviewer runs in CI on every pull request, ahead of the human pass.
Delivery: an agentic pipeline connects the ticket queue to GitHub - plan, branch, implement in an isolated worktree, open a draft PR - with three human approval gates between intent and merge.
Knowledge: an LLM-maintained knowledge base, including a pipeline that converts Oracle CCS reference documentation into validated training data for Bedrock.
Outcome
- Alert-to-fix-PR and ticket-to-draft-PR run without an engineer driving - engineers review instead of type.
- Every merge still has a named human approver; the state machine makes each step auditable.
- Packaged as config-driven tooling the whole team runs, not a personal script.