Agentic AI Engineer - Senior Manager

Location:

Anywhere in Country

At EY, we’re all in to shape your future with confidence.

We’ll help you succeed in a globally connected powerhouse of diverse teams and take your career wherever you want it to go. Join EY and help to build a better working world.

Build AI agents that carry meaningful work from intent to a verified outcome, not merely produce an impressive response. You will engineer the execution loops, tools, context, permissions, evaluations, and runtime controls that let agents operate across complex workflows while remaining understandable, steerable, and safe when systems or assumptions fail.

This role sits where applied AI, distributed systems, and product engineering meet. You will work on long-running behavior, real enterprise integrations, and difficult production failures, deciding when a deterministic workflow is enough, when autonomy adds value, and whether each added degree of freedom earns its operational and user-experience cost.

Challenger is a new EY program focused on building the firm’s next generation of services. It pursues opportunities where AI enables a fundamentally different service and way of delivering it, including rethinking existing offerings and developing new ones.

Small, dedicated teams build and test ideas in real conditions. Solutions that demonstrate clear business value and operational readiness transition to the business for adoption and scale.

Challenger Core develops and maintains the shared technology, methods, and risk and quality guardrails that support this work. Each team contributes reusable tools and practices, helping make the next service faster and less costly to build.

• Choose the right execution model. Design agent architectures that use deterministic steps, a single-agent loop, or coordinated agents only when measured performance justifies the complexity.

• Build dependable tools. Create connectors and interfaces with clear contracts, validated outcomes, narrow authorization, and errors that help the agent recover or elevate.

• Engineer agent behavior. Shape planning, routing, delegation, stopping, retry, and handoff behavior across complete workflows.

• Create durable context. Combine instructions, user state, retrieval, memory, and environment feedback while preserving relevance, provenance, freshness, and permitted access.

• Bound agent authority. Implement scoped identities, credentials, budgets, approval points, and emergency containment so agents act within explicit limits.

• Evaluate complete trajectories. Build representative tests that measure verified task success, tool use, recovery, human takeover, and prohibited actions rather than isolated responses.

• Operate long-running execution. Use durable state, checkpointing, safe retries, versioned releases, rollback, and graceful degradation to keep work recoverable.

• Make behavior observable. Instrument end-to-end traces and replay so teams can understand failures, diagnose patterns, and turn production evidence into improvements.

• Protect user control. Design clear intervention points so people can inspect, interrupt, redirect, and safely resume agent work.

• Build shared capability. Partner across product, platform, knowledge, security, and evaluation teams and publish reusable agent patterns and operational lessons.

• Production engineering:
  Strong software engineering and distributed-systems judgment, with practical experience building services that remain dependable under failure.

• Agent architecture:
  Ability to reason about planning, tools, state, handoffs, and coordination while choosing a simpler workflow when it performs better.

• AI application depth:
  Practical knowledge of model APIs, structured…