Data Scientist – Generative AI

This is a fully remote position, open to applicants in Brazil.

📋 Description

• Design the architecture of the cognitive engine, encompassing the API input payload to the output JSON.

• Develop specialized prompts that incorporate chain-of-thought reasoning, anti-hallucination safeguards, and JSON Schema.

• Execute multimodal analysis utilizing Gemini, which includes OCR and computer vision for documents, structured field extraction, and identifying inconsistencies.

• Organize RAG through semantic chunking, metadata based on coverage/clause/limit, embeddings, hybrid search, and reranking.

• Leverage historical cases to fine-tune the model using few-shot learning, focusing on edge cases.

• Establish an evaluation framework that includes automated test suites, benchmarks for accuracy, precision, and recall, as well as regression testing.

• Oversee inference quality, costs, and latency; implement LLM observability and manage prompt versioning.

• Confirm results in collaboration with business analysts.

⛳️ Requirements

• Proficiency in advanced Python and REST APIs (FastAPI or similar).

• Practical experience with LLMs in production settings, including prompt engineering, structured JSON Schema output, function calling, and anti-hallucination methods.

• Demonstrated expertise with RAG, encompassing embeddings, vector databases (pgvector, Vertex AI Vector Search, Pinecone, Qdrant, or similar), chunking, reranking, and retrieval assessment.

• Experience with multimodal models (preferably Gemini) and OCR/computer vision for documents and images.

• Familiarity with LLM evaluation methodologies, including golden sets, accuracy/precision/recall metrics, and regression testing.

• Knowledge of Databricks (notebooks, MLflow, Delta Lake) and AWS to collaborate effectively with data engineering teams.

• Proficiency with Git, adherence to coding best practices, and automated testing.

• Nice to have: Experience with Google Cloud / Vertex AI (Gemini API); LangChain, LlamaIndex, or LangGraph; Ragas, DeepEval, or LangSmith; familiarity with insurance, claims, claims adjustment, or legal and contractual documents; MLflow; understanding of Brazil’s LGPD and data anonymization; LLM guardrails and red teaming; certifications in Databricks, Google Cloud, or AI.

🏝️ Benefits

• Anywhere Office: Geographic flexibility to work from wherever suits you best.

• Partner discounts: Access to coupons and referral offers from partners for car rentals, language schools, undergraduate and graduate programs, and more.

• Talent referral program with rewards.

• Day off for your birthday.

• TotalPass/Wellhub.

• Telemedicine platform offering integrated physical and mental healthcare services.

• Health and quality-of-life initiatives.

• Relaxation spaces in our offices, equipped with bean bags, a pool table, video games, a break room, a fully equipped kitchen, and coffee.