Role summary: Builds the agents that make the migration trustworthy — parity validation, secrets/PII scanning, and the governance/PR layer — plus the Migration Co-Pilot that supports Suncorp app teams.
Key responsibilities
- Build the Parity Validator: trigger old vs new pipelines on non-prod, compare artefacts (binaries, image digests, test outcomes) rather than exit codes, and post structured diffs as PR comments.
- Build the Secrets & PII scanning agent (regex/entropy detection with LLM triage), with masked findings routed to owning teams and hard blocking gates for high-severity hits.
- Build the Governance/PR agent — the single choke-point for all GitHub writes: templated PRs, CODEOWNERS review requests, labels mirroring migration state, idempotent operations.
- Build the retrieval-grounded Migration Co-Pilot chat (grounded strictly on golden-path docs and per-repo migration notes; no invention).
- Use GitHub Copilot throughout; maintain evaluation suites that measure agent accuracy and manual-intervention rates — the numbers reported to Suncorp.
Required skills & experience:
- 5+ years software engineering (Python preferred) with solid API and automation background.
- LLM application development with an emphasis on guardrails: structured outputs, grounding/retrieval, evaluation, hallucination control. Daily fluency with GitHub Copilot.
- Security-adjacent engineering sensibility: secret detection, log redaction, audit trails; understands why raw secrets must never reach a model or a log.
- CI/CD and Git fluency sufficient to reason about build artefacts and parity.