ABOUT THE ROLE - The Agentic Engineer / AI Orchestrator is the senior technical role on a delivery pod - someone who can work across frontend, backend, and database, who directs AI coding tools to do most of the production code writing, and who is the last line of defence when something breaks in production. You will not be writing most of the code yourself. AI coding tools will do that under your direction. Your value is in deciding what to build, reviewing what the AI produces, and debugging the complex cross-system issues only an experienced engineer can solve.
>
WHAT YOU WILL DO - Direct AI coding tools to produce production-quality code across frontend, backend, and database layers.
>
- Review and validate every piece of AI-generated code before it ships. Catch logic errors, missed edge cases, performance traps, and security issues that the AI does not see.
>
- Configure and tune AI code-review tools to flag the right issues for human attention.
>
- Debug complex production issues that span multiple
>
- Hold architecture accountability for your pod - catch design problems before they reach production.
>
- Mentor more junior engineers on AI tool fluency.
>
MUST-HAVE QUALIFICATIONS AI-native engineering practice (mandatory)
The candidate we want isn't on the way to this - they're already there. This role is about bringing in someone who has lived that workflow long enough to have sharp, specific opinions on where today's tools fall short. In practice, that means you can put the following in front of us:
- Something you've actually shipped. Point us to a real project - in production, a side project, open source, or a client engagement - where agents produced the bulk of the work while you steered. Give us a live URL or a repo we can open and read. We'll want you to talk us through where the agent delivered, where it stumbled, and, most importantly, how that shaped the changes you made to your harness or setup - not just tweaks to the final output.
>
- A genuine specs-and-evals habit. Specs come before code; evals come before you sign off on anything as complete. To be clear, this isn't about working off a requirements document someone hands you - it's a discipline you've built and own yourself. If the truthful answer is "I gave it a shot once, then drifted back to vibe coding," this isn't the fit.
>
- Real command of the autonomy slider. Given any particular change, you can explain why it belongs in one bucket or another - human-written, agent-suggested, agent-shipped-with-review, or fully autonomous - and you make that call intentionally, every time. "The agent did it" doesn't excuse shipping the wrong thing.
>
- Modern engineering hygiene - Git, code review, CI/CD (Continuous Integration / Continuous Deployment), automated testing, observability.
>
Spec Discipline You can write a spec that an agent can act on and that a second Member of Technical Staff can sign off on without needing a verbal walkthrough. That means:
- Sharp goals and non-goals - what this work is meant to achieve, and just as importantly, what it deliberately leaves out of scope.
>
- Acceptance criteria framed as observable behaviour - what the system should do, not hints about how to build it.
>
- Observability and audit-log expectations stated upfront, designed in from the start rather than bolted on after the fact.
>
- Compliance and regulatory implications named explicitly wherever they apply, never left implicit.
>
- Linked skills and evals that the implementation is expected to use or build on.
>
- Bachelor's degree in Computer Science, Engineering, or equivalent practical experience.
>
AI Tool Knowledge and Fluency (Required) - must demonstrate at least 12 months daily use
- Daily active use of at least TWO of the following AI coding tools on production work in the last 12 months: GitHub Copilot, Cursor, Claude Code, Windsurf (formerly Codeium), Aider, Continue.dev, Sourcegraph Cody, Tabnine, Replit Agent, JetBrains AI Assistant.
>
- Comfortable with AI-driven code review - has used GitHub Copilot Code Review, CodeRabbit, Greptile, Qodo (formerly Codium), or similar tools, AND understands their limits.
>
- Has worked with AI agents for coding - i.e., tools that take a task and execute multiple steps (write code, run tests, fix errors, submit pull request) autonomously. Examples: Claude Code in agent mode, Cursor's Agent mode, Devin, OpenHands, Aider in architect mode.
>
- Comfortable writing effective prompts for code generation - including providing context files, defining constraints, and specifying output formats.
>
- Understands the concept of MCP (Model Context Protocol) - the standard for connecting AI tools to external data sources, codebases, and tools. Bonus if they've used or built MCP servers.
>
- Can describe specific failure modes of AI-generated code they have encountered - security issues (e.g., hardcoded secrets, SQL injection in generated queries), performance issues (e.g., N+1 queries, inefficient loops), and architectural issues (e.g., violating existing patterns).
>
- Understands prompt injection as a security risk and knows how to mitigate it in AI-augmented systems.
>
- Familiarity with evaluation ("eval") frameworks for AI output - knows that production AI features need automated tests, not just manual review.
>
Information Security Responsibilities- Promote and enforce awareness of key information security practices, including acceptable use of information assets, malware protection, and password security protocols
- Identify, assess, and report security risks, focusing on how these risks impact the confidentiality, integrity, and availability of information assets
- Understand and evaluate how data is stored, processed, or transmitted, ensuring compliance with data privacy and protection standards (GDPR, CCPA, etc.)
- Ensure data protection measures are integrated throughout the information lifecycle to safeguard sensitive information
We invite you to stay connected with us by subscribing to our monthly job openings alert here.
#BI-Remote#LI-Remote 
