Summary :

Build AI native, data centric products on SAP BTP Datasphere by combining strong enterprise data warehousing and semantic modeling expertise with agentic AI architectures (LLMs + tools + retrieval + evaluation).

The focus is to move beyond dashboards into intelligent data experiencesdata agents, conversational analytics, and grounded insightsbuilt on governed Datasphere models and integrated enterprise sources.

SAP Datasphere is positioned as a data warehousing solution with integration capabilities.

Core Responsibilities :

1) AI Native Data Product Engineering (on Datasphere):

- Design and implement governed data products using Datasphere concepts such as Spaces and shareable models/views, enabling teams to explore, transform, and share curated datasets across domains.

- Build semantic models that are fit for both analytics and AI consumption (clear entity definitions, measures, hierarchies, lineage-friendly design).

2) Retrieval + Grounding (RAG) over Enterprise Data :

- Create grounded AI experiences by connecting LLM applications to Datasphere s curated models and enterprise sources (SAP and non SAP), ensuring responses are traceable to governed data.

- Engineer retrieval strategies that respect domain boundaries (spaces), freshness needs, and access controls, so AI outputs remain reliable and compliant.

3) Hybrid Modernization & Migration (BW bridge patterns) :

- Enable transition paths from legacy warehouse investments by leveraging approaches such as reusing SAP BW models and skills with Datasphere / BW bridge, supporting phased cloud modernization.

4) Lakehouse style Layering & Data Quality by Design :

- Implement layered design patterns (e.g., Bronze/Silver/Gold) to land raw data, cleanse/validate, and publish analytics ready modelswhile maintaining clear rules for what s exposed for consumption.

- Embed quality controls, validation checks, and reproducible transformations as part of the delivery lifecycle.

5) Agentic Orchestration & Tooling :

- Build data agents that can plan, call tools (query/metadata/lineage), retrieve context, and generate answers with citationsbacked by deterministic checks and fallback behaviors.

- Implement prompt templates, tool schemas, and safe action boundaries for enterprise-grade usage.

6) Evaluation, Observability & Responsible AI ":

- Establish offline/online evaluation loops (golden questions, regression suites, behavior tests) for conversational analytics and data agents.

- Add telemetry for AI interactions (latency, grounding rate, failure modes) to improve reliability and cost efficiency.

7) Integration & Collaboration :

- Partner closely with business, data governance, and platform teams to align data products with real decisions and operational workflows.

- Drive reusable patterns and accelerators for repeatable delivery across domains.

Primary Skills :

- (AI Native Must Have) SAP BTP Datasphere : data modeling, spaces, sharing patterns, enterprise semantic design.

- Strong data warehousing fundamentals and ability to translate business domains into governed analytical models.

- Hands-on building with LLMs + RAG (retrieval, grounding, prompt/tool design, evaluation).

- Solid software engineering fundamentals : testability, CI/CD mindset, reliable integration.

Secondary / Strongly Beneficial Skills :

- Migration/modernization experience leveraging BW bridge style transition patterns. Layered architecture implementation (Bronze/Silver/Gold) for scalable analytics delivery.

- Familiarity with vector search / embedding pipelines (when integrating external AI retrieval components).

- What This Role Does Not Center On Training foundation models from scratch (the emphasis is on building agentic apps and governed retrieval on enterprise data).

- AI assisted only delivery this role owns the AI behavior (grounding, evaluation, safety) end to end.

- Value Delivered Faster path from data to decision through conversational + agentic analytics grounded in governed Datasphere models.

- Scalable modernization of hybrid data estates via patterns like BW bridge. Higher trust AI outputs by implementing layered quality + evaluation loops.