Cientista de Dados Sênior - IA Generativa e Agentes

This position is listed on behalf of a partner company, who manages all applications and next steps. Our partner is looking for a Cientista de Dados Sênior - IA Generativa e Agentes based in Brazil. As a Senior Data Scientist, you will lead the intelligence layer behind a new generation of AI agents designed to automate complex back-office processes. You will work on the development of five agents supporting areas such as store openings, regulatory intelligence, sanitary compliance, purchasing, and health-plan registration. The role combines Generative AI, LLMs, RAG, document intelligence, evaluation frameworks, and AI security in a highly practical environment. You will collaborate closely with solution architecture, development, and business teams while working with sensitive data and enterprise systems. Governance, LGPD compliance, traceability, and responsible AI will be embedded throughout the project lifecycle. This is a remote, hands-on opportunity to shape reliable, scalable, and business-focused AI solutions.

Develop, version, and continuously improve system prompts for AI agents, defining their roles and scope, guardrails, permitted actions, SLAs, long-running process flows, tool-calling policies, output formats, and human escalation mechanisms. Prepare and structure knowledge bases for RAG applications, working with regulatory standards, internal policies, RFPs, and operational rules while applying document cleaning, sectioning, metadata, validity periods, categories, and access controls. Evaluate retrieval quality through metrics such as recall@k and access-group isolation, continuously refining RAG strategies to improve relevance, accuracy, and security. Build OCR and information-extraction workflows for PDFs, photographs, and handwritten documents, establishing quality metrics and appropriate handling for illegible or incomplete materials. Implement data anonymization and masking services, as well as input guardrails against prompt injection and output controls covering schemas, data leakage, and prohibited actions. Build golden datasets and prompt evaluation harnesses with regression testing across versions, measuring accuracy, quality, and cost on a case-by-case basis. Evaluate and validate LLMs from different providers according to specific use cases, applying AI FinOps practices such as model routing, batch processing, and prompt caching to balance performance and cost. Build and configure AI agents in collaboration with the technical team, supporting integrated testing, user acceptance testing, and assisted operations. Document technical decisions and contribute to architecture committee materials, including assessments related to cost, security, data handling, governance, and solution design. Requirements: Bachelor’s degree in Computer Science, Statistics, Mathematics, Engineering, Information Systems, or a related field. At least 4 years of professional experience in Data Science or Machine Learning, including hands-on experience with Generative AI projects. Practical experience with LLMs, including prompt engineering, function calling and tools, structured outputs using JSON Schema, and AI agent development. Strong experience with RAG architectures, including embeddings, vector databases, hybrid search, reranking, and retrieval-quality evaluation. Proficiency in Python for experimentation, evaluation, data preparation, and development of AI-focused solutions. Experience building datasets and evaluation metrics for LLM-based solutions, including accuracy, precision/recall, and LLM-as-a-judge approaches. Knowledge of OCR and document information extraction using multimodal models and/or Document AI technologies. Knowledge of AI governance and LGPD requirements, including data masking, anonymization, and data minimization practices. Intermediate to advanced English communication skills. Experience with Azure OpenAI and/or Microsoft Foundry and MongoDB Atlas Vector Search is considered a strong advantage. Knowledge of the Model Context Protocol (MCP) and agent patterns such as ReAct and stateful workflows is desirable. Experience with LLM security practices, including prompt injection, red teaming, and guardrail implementation, is a plus. Knowledge of AI FinOps, including token-cost estimation, optimization, and comparison of models and providers, is desirable. Familiarity with Node.js and TypeScript is an advantage. Experience with RPA or integrations involving ServiceNow, Microsoft Teams/Graph, or SAP is valued. Knowledge of healthcare, regulatory environments such as ANVISA, procurement, or HR benefits is considered a plus. Cloud or AI certifications, such as Azure AI Engineer certification, are desirable. Benefits: Fully remote work model, allowing you to contribute to a technically advanced AI project from anywhere in Brazil. Five-month project duration, with the possibility of extension or internalization. Opportunity to work directly with Generative AI, LLMs, RAG, AI agents, OCR, document intelligence, and enterprise automation. Hands-on experience developing AI solutions for complex business processes involving sensitive and regulated data. Exposure to advanced AI governance, LGPD, security, evaluation, traceability, and responsible AI practices. Opportunity to collaborate with a multidisciplinary squad including solution architecture, software development, and business analysis. Experience with enterprise AI platforms, model evaluation, AI FinOps, integrations, and production-oriented agent workflows.