Agentic AI Engineer)

About this role

Catapult seeks an Agentic AI Engineer to design and deploy production-grade multi-agent AI systems that provide trustworthy, calibrated performance insights for coaches and athletes across professional sports. You'll build specialist agents with confidence calibration, human escalation awareness, and domain knowledge injection—focusing on systems that earn trust rather than assume it.

What you'll do

• Design and build specialist AI agents that reason over different dimensions of athletic performance and compose their outputs into unified recommendations
• Develop workflow engines that encode domain scientist expertise into validated, production-ready agent skills with version control and regression testing
• Implement confidence calibration, evaluation frameworks, and decision intelligence layers that transform agent outputs into trustworthy, escalation-aware recommendations
• Build human-in-the-loop architectures with escalation models and consequence classification to ensure practitioners retain decision ownership
• Create knowledge acquisition workflows (annotation interfaces, review queues) that enable rapid, validated skill updates across multi-agent systems
• Establish observability, drift detection, and evaluation harnesses for production agentic systems running against full input distributions

What they're looking for

• Production agentic AI systems (memory, tool use, multi-step reasoning, calibrated outputs)
• Multi-agent frameworks and orchestration (dependency routing, specialist composition, response synthesis)
• Confidence calibration and evaluation (Platt scaling, isotonic regression, ECE, evaluation harnesses)
• Production RAG with reranking and retrieval quality optimization
• Foundation model fine-tuning and domain-specific adaptation
• Python and Golang; LLM observability and drift detection
• Human-in-the-loop architectures and escalation modeling
• AWS infrastructure (ECS, EC2, Lambda, SNS, SQS), GraphQL, REST, gRPC, Postgres, Mongo