Principal Data Scientist, AI & Enterprise

Job Description

Teach AI how business actually works

AI agents are starting to take on real business work: releasing blocked invoices, re-planning supply, preparing the financial close. To do it safely, they need more than access to data. They need to know what an order, a supplier and a cost center are, how they connect, which rules and authorizations apply, and what happened the last time someone made a similar decision.

When executing mission-critical business processes, an answer or action that is right 90% of the time is not good enough.

SAP systems hold the context for many of the world's largest companies. SAP's knowledge graph for SAP S/4

HANA alone is based on 452,000 tables, 80,000 CDS views (semantic data models) and 7.3 million fields. Yet much of the meaning, especially in each company's own extensions, configuration and decision history, is still implicit. A 'customer' in sales, a 'business partner' in finance and an 'account' in CRM can be the same company described three ways, and an agent deciding whether to raise a credit limit has to know that.

This is not a data-cleaning problem, it is a knowledge-representation problem. Today that context, the semantic layer agents rely on, is mostly modeled by hand, one company at a time. We think it can be learned, so AI agents can act on business data, not just read it. SAP's Autonomous Enterprise depends on that, and it is what makes the agents people work with in Joule Work, SAP's AI workspace, accurate enough to trust.

As Principal Data Scientist, you will define and lead our research toward that goal and work with product teams to ship the results. This is a research and modeling role, not product analytics.

What you will work on

• A research agenda with a point of view. Pick the two or three bets where SAP can lead in AI on business data, agree up front how we will know each one is working, and stop the ones that aren't. Closely collaborate with top universities and institutes such as Stanford HAI.
• What comes next. Track emerging research, such as relational and graph foundation models, neuro-symbolic agents and context graphs, test it on SAP data before the market settles, and make the case for the next bet.
• Business semantics that can be learned. New methods to learn, align and evolve ontologies and knowledge graphs of business objects, rules and actions from SAP metadata, master data, documents and process data, combining LLM-assisted ontology learning, schema matching and entity resolution with targeted expert review.
• Agents grounded in business semantics. Graph-grounded retrieval; semantic parsing into SQL, SPARQL and OData calls, including finding the right API, across very large schemas; and governed actions: checks that an agent's plan respects business rules and the user's authorizations before it runs.
• Models that combine structure and statistics. Graph learning and relational foundation models (such as SAP-RPT-1.6), using knowledge-graph structure to predict, explain and simulate business outcomes.
• The yardstick. Benchmarks and simulation environments for agents on ERP data, where today's public benchmarks fall short, so SAP knows when AI is accurate and reliable enough to act on.
• Results that land. Work directly with data teams and domain experts in co-innovation projects, turning messy enterprise data problems into tractable ontology problems, and work with the teams behind SAP Knowledge Graph, SAP Business Data Cloud and Joule, SAP's AI assistant, to ship what works.
• A stronger research community. Publish at top venues, release code and benchmarks where we can, co-lead university collaborations, and mentor scientists and PhD interns.

In your first year, success could look like:

• Two or three research directions you defined, agreed with leadership, at least one with a product partner and one longer-horizon
• A result running in an SAP product or pilot
• Papers submitted, and a benchmark or open-source release others use
• Scientists doing better work because of you

What you'll bring

Required

• Typically 8+ years in applied research, data science or machine learning, or a PhD and 5+ years, including research you led that changed a product or platform. An MS or PhD in computer science, machine learning, statistics, mathematics or a related field, or equivalent practical experience.
• Deep expertise in one of these areas and hands-on experience in another, with at least one of the two in knowledge graphs, ontologies or semantic data integration:
  • Knowledge graphs,…