Job Description

Start your career where innovation meets real impact.

Sapiens Accelerate is our global Early Careers Programme, designed for graduates with up to 18 months of experience, who are ready to learn, contribute, and shape the future of insurance. From day one, you'll work on meaningful projects, collaborate with experts across the business, build practical AI skills, and develop the capabilities needed to thrive in a global technology company.

The role

Agentic Data Foundation is the action layer. AI agents reason over governed data products and ML insight, not raw tables. We expose data and ML as governed MCP tools, build Synapse (our insurance ontology and semantic knowledge layer) and context as a service. Our agents are read-only, bound to one customer, fully logged and approved by people. This is a building role on a greenfield platform, with short prototype cycles and direct impact on product direction.

Our stack: Python for agentic workloads, MCP, LLMs (Azure OpenAI), GraphQL APIs, Azure Databricks and Unity Catalog, ontology standards (SKOS, OWL, RDF and W3C).

Responsibilities

What you will do

• Build MCP tools. Expose governed data and ML outputs as typed, policy-checked, read-only tools for agents.
• Prototype agents. Build multi-step, tool-using agent workflows for insurance cases such as claims triage and lapse risk.
• Evaluate and secure. Write evaluation harnesses for reliability, plus guardrail tests (tenant isolation, free text treated as data, every call logged).
• Add meaning. Help build ontology, knowledge-graph and context-service components with the Synapse team.
• Ship and explain. Use AI coding tools daily, document and demo every prototype.

Requirements

You should have

• Degree in Computer Science, Engineering, AI/Data Science or similar; 0–18 months of experience (internships and projects count).
• Python, plus hands-on LLM projects: prompting, structured output, function/tool calling and handling failures.
• SQL and API basics (REST or GraphQL); Git.
• AI coding tools (Claude Code, Cursor, Copilot) in your daily workflow.
• Self-directed: comfortable with ambiguity, you make decisions and you ship.

Even better if you have

• LangChain/LangGraph or similar; MCP or other agent standards.
• Embeddings, vector databases and RAG in a real application; building evaluation frameworks for LLM outputs.
• Knowledge graphs or ontologies (RDF, OWL, SPARQL); TypeScript.

Benefits

Why join

You will design agent architectures on governed enterprise data, solve reliability and evaluation problems, and see your work in front of insurers. Fast path to AI engineer or agent architect.