Data Scientist

Description

✨ Our mission here at Mundiale is to simplify the relationship between brands and their consumers, providing more human, assertive, and fluid interactions through cutting-edge technology and our own methodology. To do this, we are looking for many skilled people to join our team of Munders and live, every day, new and exciting challenges! We have an opportunity for a **Data Scientist**! Check out this opportunity below and come ramp up your career in a company that never stops growing! **ABOUT THE POSITION** You will build and operate the AI agents that serve, sell, and resolve for the brands that are our clients, from designing the multi\-agent architecture to what runs in production interacting with the end consumer. **RESPONSIBILITIES AND ATTRIBUTES** * Design and implement agents and multi\-agent architectures (orchestrator plus specialists), defining the scope of each agent, routing rules, and handoff. * Write and refine system prompts as an engineering artifact, versioned, with decision history and validation before deployment. * Structure knowledge bases (RAG), taking care of content slicing and retrieval quality. * Build and run test and evaluation suites for agents (end-to-end cases, edge scenarios, regression, LLM as judge). * Instrument and analyze production metrics such as resolution rate, reasons for escalation, conversion, cost per conversation, latency, and token consumption, using this to decide the next improvement. * Choose model, effort, and temperature per agent with an explicit trade\-off between quality, cost, and latency. * Translate technical results into clear business reading for product teams and for the client. * Work together with product, engineering, and operations to prototype, validate, and deploy solutions. **REQUIREMENTS AND QUALIFICATIONS** * Minimum experience in building AI agents. * Proven experience writing and maintaining agent system prompts. * Solid experience in Python. * Practical experience with LLMs in real applications, going beyond chat usage, with tool calling, structured output, and context control. * Knowledge of RAG and semantic representation (embeddings, chunking, retrieval quality). * Git and good practices of versioning and review. * Autonomy to investigate non-deterministic behavior with method, starting from hypothesis, evidence, correction, and verification. **DIFFERENTIALS** * Experience with agents in production in conversational channels (WhatsApp, chat, voice) and with the design of escalation to human service. * Guardrails, intent classification, and response security and containment mechanisms. * Observability and evaluation of LLM (LangSmith, Langfuse, MLflow, Weights \& Biases or equivalent). * Familiarity with MCP and with agent orchestration frameworks. * Analysis of conversational data (Pandas plus visualization) to extract failure patterns from real volume of services. * Notions of AI FinOps, reducing cost per conversation without losing quality. * Interest in AI applied to product and automation of conversation flows. **ADDITIONAL INFORMATION** AND YOU HAVE: Flexible Benefit (you define how to use it according to your needs: culture, transport app, subscriptions, gym, food, etc.) ️ Unimed Unipart Health Plan Career Path Remote Work Day Off Wellhub ️ Short Friday Collaborative and dynamic work environment