Senior AI/ML Engineer - Agentic AI

Summary 

We're building specialized foundation models and AI agents that accelerate SAP customers' data transformation journeys. 

The agents you build will directly power SAP's Autonomous Enterprise, where AI runs core business processes end-to-end across finance, supply chain, HR, and procurement at global scale. 

You'll set technical direction, define how we architect and scale multi-agent systems, and raise the engineering bar across a global team. You'll work directly with pretraining and fine-tuning team leads in Europe, India, and early-adopter customers. 

We want someone who has shipped agentic systems in production and knows where they break. 

What you'll do 

• Architect and lead multi-agent systems: design, orchestration patterns, failure modes, memory, planning, and human-in-the-loop 

• Own the path from prototype to production: containerization, guardrails, cost and latency optimization, scalable serving 

• Define the team's evaluation strategy: offline/online harnesses, trajectory quality, tool-call accuracy, regression testing, CI/CD eval gates 

• Lead instrumentation and observability: tracing, span capture, automated scoring, closing the trace → eval → fix loop 

• Drive tool integration architecture via MCP across multiple product teams 

• Mentor junior and mid-level engineers through code and architecture reviews; set engineering standards 

What you bring 

Education: BS, MS, or PhD in Computer Science, ML, or a related field. 

Experience: 6+ years building and shipping ML systems, with 3+ years hands-on with LLMs and agents in production. 

Core technical skills 

• Expert Python; strong fundamentals: system design, testing, modularity, async, API design 

• PyTorch; working knowledge of fine-tuning and PEFT methods (LoRA, QLoRA) 

• LLM application development: prompting, structured outputs, tool calling, context management 

• Inference optimization: vLLM, TensorRT-LLM, quantization (int8, int4, GPTQ, AWQ) 

• Human-in-the-loop annotation workflows at scale 

Agent frameworks and orchestration (production experience with at least three) 

• LangGraph / LangChain 

• CrewAI, AutoGen/AG2, or equivalent 

• MCP (Model Context Protocol) 

• Coding agents: Claude Code, OpenCode, or similar 

Evaluation and observability (production experience with at least two) 

• Langfuse, LangSmith, Arize, or equivalent 

• LLM-as-judge evaluators, CI/CD eval gates 

Leadership 

• Demonstrated track record ment