Overview
In this role you design end-to-end AI architectures for insurance pricing, building agentic workflows and production deployment. You work within the Global P&C; Pricing team to embed GenAI capabilities across the pricing lifecycle, collaborating with actuaries, data scientists, and AI architects. You translate complex AI concepts into actionable insights for senior stakeholders and contribute to standards for AI architectures. This position offers the chance to shape pricing tools and regulatory documentation with scalable, robust AI solutions. You will operate at the intersection of actuarial science, data engineering, and cutting-edge AI to drive technical excellence.
Responsabilità Design and develop end-to-end AI architectures for insurance pricing, including retrieval systems and production deployment
Integrate AI/ML models (GLMs, LightGBM) and rating engines into agentic pricing architectures for natural-language interaction
Build and maintain LLM-powered tools for tariff analysis, pricing code generation, simulations, and regulatory docs




Prototype AI solutions rapidly and present findings to senior stakeholders
Collaborate with actuaries, data scientists, and AI Transformation Architects to ensure scalable, robust solutions aligned with business goals
Contribute to AI architecture standards and best practices within Global P&C; Pricing
Requisiti fondamentali 3–5+ years designing and deploying end-to-end GenAI solutions
Hands-on experience with end-to-end AI architectures, LLM-powered tools, and model integration (GLMs, LightGBM)
Proficiency in Python; SQL, ETL, and data modeling familiarity
Experience with Cloud Services (AWS, GCP, Azure)
Experience with LLM APIs (OpenAI, Azure OpenAI, Anthropic, HuggingFace)
Experience with agentic and RAG frameworks (LangChain, LlamaIndex, AutoGen) and vector databases (pgvector, Weaviate, Pinecone)
Familiarity with MLOps and model serving concepts (MLflow, CI/CD)
Git, Docker, and Kubernetes knowledge
Fluent in Itali

📌 AI ARCHITECT (Milano)
🏢 Allianz
📍 Milano