Senior Software Engineer-AI/Agent Engineer & Tester

Position Description:

Your future duties and responsibilities:

• Responsibilities Agent Implementation Implements LangGraph agent nodes and graphs to architectural design specified by APEX Lead. Builds and wires routing logic. No node ships without a unit test; no graph ships without an integration test. Agent Testing - Unit Level Every LangGraph node tested in isolation with mocked state. Input/output contracts verified independently of the full graph. Non-deterministic behavior is controlled via mocked LLM outputs or seeded inputs - flaky tests are not acceptable. Agent Testing - Graph Level Full end-to-end agent runs with controlled inputs — routing decisions, state transitions, output fidelity, and failure modes all covered. Agent Regression Suite Responsible for the regression harness so prompt changes, model version bumps, or node refactors don't silently degrade behavior. Tied into Azure DevOps CI so every PR runs the suite. Responsible for suite health - flaky, skipped, or stale tests are actively managed. Bug Fixing Ownership When an agent bug is reported, Responsible for the full cycle: reproduce, isolate, fix, write the regression test that would have caught it, ship. No fix is complete without a corresponding test added to the regression suite. Playwright E2E Coverage Responsible for the Playwright test suite across the entire GUI frontend. APEX Lead will specify which user flows must be covered; This role builds, maintains, and actively tracks coverage gaps. Wired into the Azure DevOps pipeline on every PR. Adversarial Agent Testing Prompt injection, cross-submission data leakage, role-boundary violations, malformed intake inputs. Dedicated adversarial suite calibrated to bank-grade governance platform requirements. Prompt Versioning and Management Structured approach to prompt storage, change review, and rollback - a lightweight prompt registry tied to regression test runs so prompt changes are treated with the same discipline as code changes. Production Agent Observability LangSmith tracing or equivalent - node-level latency, routing decision logs, token usage, and failure instrumentation in production. Document Intelligence Agent Layer When apex-documents lands, responsible for the AI extraction pipeline - structured field extraction from uploaded documents - coordinating the interface contract with backend service. Full test coverage from initial implementation. Model Version and Cost Management Vertex AI model pinning, token usage baselines per workflow, cost-per-request tracking. Surfaces these metrics on every regression run so model upgrades don't introduce silent cost or quality regressions. GCP Infrastructure and Terraform Cloud Run, Cloud SQL, IAP, Hashicorp Vault - responsible for infra changes to the Terraform module structure. APEX Lead will review and approve; This role will execute and maintain.