Developer

About the company

Where you belong You’ll join Applied AI Builders, a small, hands-on pod that works closely with the CTO for Industry AI ENR. Our job is to take ambiguous problems, half-formed ideas, and evolving initiatives, and turn them into prototypes, proofs, and production-ready outcomes. Some weeks we’re embedded with a product team, partnering on a specific AI mission. Other weeks we’re stress-testing a raw idea to see whether it deserves to be built, adapted, or respectfully retired.

Your tasks

What you'll build We move across the prototype-to-production spectrum: we build fast, validate honestly, and partner with platform and product teams to scale what proves valuable. The stack varies —

• Python
• TypeScript
• Terraform
• BTP
• Joule Studio
• CAP
• Fiori

— whatever the mission needs — but we care far more about your taste for good code, your ability to supervise Claude-style code generators and AI brainstormers, and your architectural judgment than about fluency in any single language.

You’ll use tools like Claude as accelerators, but you’ll read what they produce, challenge their choices, and know when the AI is making the wrong call — both in code and in idea generation.

You’ll own modules and prototypes end-to-end, pair with other builders in the pod, and spend real time embedded with product teams on focused AI missions.

The pod’s missions sit at the frontier of AI-native engineering. One current mission: ontologies. We’re designing and evolving ontologies for our domains, building agents on top of them, and proving — with data — how ontology-grounded agents compare to raw prompt-and-MCP-based agents. We’re defining the evaluation methodology that makes that comparison honest. This is one mission of many; more will follow as we double-click into AI engineering.

Requirements

• 3 - 6 years of hands-on experience with LLM APIs, at least one agentic orchestration framework, RAG pipelines, prompt engineering, and evals (human, LLM-as-judge, automated), plus a strong sense for good vs. bad code.
• You use Claude-style tools to generate code and ideas, but you read and challenge their output. Comfort with Python helps; enjoyment in learning new stacks (BTP, CAP, HANA Cloud, Joule, TypeScript, Terraform) matters more.
• Learning agility (non-negotiable, and the most important value in this pod) — you rapidly pick up new models, frameworks, SAP technologies, ontologies, and industry processes. You enjoy updating your mental model as the landscape shifts, which it will, frequently.
• Bias for action — you move quickly from ambiguity to a working prototype and learn through execution.
• Critical thinking — you ask the right questions, so we stay focused on the objective instead of just managing the day.
• End-to-end ownership — you own the outcome from problem definition through build, deployment, and evaluation.