You join team Phoenix, one of the first teams adopting our AI-native SDLC, reporting to an Engineering Manager, proving it in production on real customer-facing systems. Phoenix is a cross-functional team spanning back-end, front-end, UX, testing, and product — you deliver alongside a UX designer and a Product Manager, and work with platform teams to get code to production.

• Owns and evolves the Knowledge Base end to end, from UI to backend.
• 
  
• Ships legacy-modernisation and product work through agentic workflows.
• 
  
• Builds the shared prompt and agent libraries the wider org adopts.
• 
  

What you'll do

Build agentically

• Frame and decompose problems, orchestrate agents to implement them, and verify the result.
• 
  
• Build new agentic product capabilities, and modernise pieces of the legacy Java/MS SQL platform on the modern TS/Vue/Azure stack.
• 
  

Work with agents, stay accountable

• Write implementation briefs and acceptance criteria before invoking agents.
• 
  
• Curate context aggressively — the right files, constraints, and invariants; lean on CLAUDE.md conventions, subagents, and MCP tools to scale beyond single-shot prompting.
• 
  
• Design verification before or alongside implementation: tests, runtime checks, migration validation, observability, and rollback planning — plus the evals that catch regressions automatically.
• 
  
• Handle agent permissions, context, secrets, and dependency changes with care, and keep agent-generated change sets small, understandable, and maintainable with reviewable evidence behind them.
• 
  
• Recognise when a plausible agent-generated solution is architecturally wrong, unsafe, or unnecessarily complex — and review agent-generated work before it merges. You own the outcome.
• 
  

Raise the bar

• Contribute reusable patterns, subagents, and evals to the shared libraries the wider org adopts.
• 
  
• Pair with colleagues to lift the team's overall capability — coaching is part of the job.
• 
  

Technical environment

Claude Code is our current primary agentic engineering environment: agents may plan and execute within defined constraints and governed environments, while engineers stay accountable for problem framing, architecture, acceptance criteria, security, verification, and production outcomes. Agent-generated work is backed by automated tests, runtime checks, reviewable evidence, and risk-based human approval — we match deterministic automation, structured workflows, or agents to the nature and risk of the task rather than maximising autonomy for its own sake.

• Legacy: Java, MS SQL, heritage architecture — being decomposed.
• 
  
• Modern: TypeScript, Vue, Azure, Kubernetes; cloud-native / service-oriented architecture.
• 
  
• Mechanisms: subagents, MCP tool integrations, multi-agent workflows, and shared prompt, agent, and eval libraries.
• 
  

Success in your first year

• You have shipped meaningful product and modernisation work in production via agentic workflows.
• 
  
• Your team's cycle time, lead time, and defect rate are measurably better than the baseline.
• 
  
• You have contributed reusable patterns, subagents, or evals to the org-wide libraries.
• 
  
• Engineers around you build better in part because of how you work.
• 
  
• 
  
• 
  

Qualifications

Required

• At least 5 years of expereince in Java.
• 
  
• Strong production engineer with current, daily, hands-on agentic delivery practice (Claude Code, Cursor, Aider, or equivalent).
• 
  
• Experience in at least one of: TypeScript / Vue, Azure-hosted systems.
• 
  
• Deep software-engineering judgement independent of any tool — strong testing instincts, comfort with operational concerns, and the ability to spot output that is plausible but architecturally wrong, unsafe, or unnecessarily complex.
• 
  
• Comfortable in ambiguity and inside legacy codebases; communicates well in writing.
•