PortBlueSky is a technology services agency that supports leading corporations in developing and delivering ambitious projects. They are seeking a Lead AI Engineer to build agentic supply-chain systems that enhance forecasting, order management, and operational decision-making by integrating AI with enterprise data systems.
Responsibilities:
- Build agentic supply-chain systems: Design, implement, and maintain agent workflows that support forecasting, order management, inventory planning, discrepancy detection, and operational recommendations
- Own agent orchestration: Build a central orchestration layer that routes tasks, coordinates multiple agents, manages tool and data-source access, and provides a clear user-facing interface
- Integrate with enterprise data systems: Connect agentic workflows to ERP data lakes, Lake House architectures, APIs, and existing AWS-hosted infrastructure
- Operate beyond prototypes: Ensure systems work on real production data and continue to perform under realistic scale, latency, reliability, and maintenance constraints
- Optimize production behavior: Monitor and improve token usage, agent routing, memory use, cost efficiency, scalability, and long-term sustainability
- Strengthen security: Implement prompt-security strategies, safe tool execution, access boundaries, and guardrails that prevent system hijacking, accidental destructive actions, or unsafe data loading patterns such as loading entire data sets into agent memory
- Use structured data contracts: Apply Pydantic models and validated input/output schemas so multi-agent systems remain predictable, typed, and auditable
- Build evaluation into the system: Create benchmarks, regression checks, reliability tests, and model-upgrade evaluations that track accuracy and behavior over time
- Lead through knowledge: Act as a senior IC, mentor, coach, and point of contact for agentic AI guidance across the project and occasionally across other teams
- Upskill the team: Help junior AI engineers and adjacent engineering teams understand agentic-system design, limitations, risks, and production practices