We are looking for an engineer who has already built with AI in production. You have shipped LLM-powered systems (agents, copilots, RAG pipelines, or AI-driven automations), you know what breaks when they meet real users, and you know how to make them reliable enough for business-critical workflows.
Your day-to-day is applied AI engineering: designing agentic workflows, integrating LLMs into internal tools and business processes, building evaluation and guardrail layers, and turning manual, high-friction workflows into AI-assisted ones that thousands of Nubankers depend on.
This is not an infrastructure role. You will not spend your days on Terraform, IAM policies, or email deliverability. Cloud fluency helps, but the core of this job is the AI layer — prompts, context, agents, evaluations, integrations — and the product judgment to know where AI genuinely helps versus where deterministic automation is the right answer.
Key Responsibilities
Applied AI & Agentic Systems
Design, build, and ship LLM-powered agents and workflows that automate complex internal processes end-to-end.
Work hands-on with frontier models and the modern AI stack: tool/function calling, structured outputs, MCP, RAG, multi-agent orchestration.
Own the full lifecycle of an AI system: from problem discovery and prototype to production hardening, monitoring, and iteration.
Evaluation & Reliability
Build evaluation harnesses, guardrails, and quality feedback loops so AI systems can be trusted in production — not just demoed.
Define what "good" looks like for non-deterministic systems and instrument it: evals, regression suites, human-in-the-loop review where it matters.
Intelligent Workflow Automation
Use orchestration platforms (e.g., n8n) and custom integrations as delivery vehicles for AI-in-the-loop automation across business units.
Integrate enterprise platforms (Slack, Google Workspace, Jira, internal APIs) into coherent, AI-assisted workflows.
AI Adoption & Governance
Drive the technical strategy for AI adoption within engineering and business workflows.
Develop governance frameworks that make AI coding assistants and agents safe, compliant, and effective — balancing developer freedom with security and operational risk.
Multiplier Work
Create Golden Paths, reference implementations, and documentation that let other teams build AI workflows safely on their own.
For Lead/IC6: act as the technical reference for applied AI in the domain, influence architecture beyond the immediate team, mentor senior engineers, and partner with ITSec and Privacy to align AI solutions with company policy.
For Senior/IC5: execute complex AI projects with high autonomy, identify workflow bottlenecks worth automating, and mentor mid-level engineers.
What are we looking for?
Must Have — Demonstrated Applied AI Experience
Shipped LLM systems in production: at least one real system with an LLM at its core — an agent, copilot, RAG application, or AI-driven automation — used by real users, not a proof of concept.
Hands-on AI engineering: practical fluency with prompt and context engineering, tool/function calling, structured outputs, and agent frameworks or orchestration patterns.
Evaluation mindset: experience measuring and improving AI output quality — evals, test sets, feedback loops — and an honest understanding of failure modes (hallucination, drift, prompt injection).
Solid software engineering foundation: proficiency in Python, TypeScript, or Clojure; strong API and integration skills; the discipline to ship maintainable systems, not notebooks.
AI product sense: the judgment to identify which problems deserve an LLM, which need deterministic automation, and which should not be automated at all.