Data Scientist/ML (Agentic AI)

Location: The role offers flexibility for occasional travel to the Warsaw office and potential international travel to Germany, approximately once a quarter.Start: Preferably ASAP or max one month notice Industry: Pharmaceuticals / Consumer Health The Mission: You are the "brain" designer. Moving beyond classic ML models, you will design complex, multi-agent workflows. Your mission is to build the cognitive architecture of our Co-pilot for sales representatives - ranging from strict Text-to-SQL routing to human-like conversational interviews - ensuring compliance and continuous improvement. Who You Are & What You'll Do: • Agentic Workflows: You have deep, practical experience building complex agent routing and state management using LangChain, LangGraph, or tools like n8n. • Multimodal & Conversational AI: You have deployed advanced RAG systems and have experience integrating TTS/STT (Text-to-Speech/Speech-to-Text) pipelines for asynchronous "Human-AI interview" conversational flows. • Text-to-SQL & Business Logic Routing: You excel at mapping natural language to exact SQL parameters. You understand that the AI shouldn't "guess" financial math; instead, it must flawlessly route intents to deterministic business logic/SQL views provided by our data teams. • Continuous Evaluation & Guardrails: You know that standard testing fails with GenAI. You will design systemic validation pipelines (e.g., LLM-as-a-judge) to monitor hallucination rates, using tools like Langfuse to establish a continuous improvement loop. • Multilingual Evaluation: You know standard English testing fails in localized markets. You will design systemic validation pipelines to evaluate RAG, intent classification, and transcription accuracy in Italian (handling medical/pharmaceutical jargon). • Optimization & Fallbacks: You understand how to optimize AI processes—utilizing prompt caching, context-window management, and configuring faster fallback models when primary LLMs time out, ensuring a seamless user experience. Must Haves: • Deep, practical experience with LangChain, LangGraph, or similar tools. • Knowledge of integration concepts for TTS/STT systems. • Strong Text-to-SQL skills for accurate data routing. • Experience with validation of generative AI and RAG pipelines. • Proficiency in Python, SQL, and Spark programming. Nice to Haves: • Familiarity with GitHub for version control. • Experience with FastAPI for application development. • Awareness of data solutions like Databricks. • Understanding of Azure cloud services.