Bolsista Mestre – Agentes, LLM, IA generativa, Prompt engineering

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

📋 Description

• Comprehend the current ecosystem of agents, including architecture, tools, logs, memories, execution flows, and human feedback mechanisms;

• Execute at least one existing agent end-to-end in a local/controlled environment;

• Map the complete execution flow of an agent and its main interfaces;

• Research the state-of-the-art in agent memory, human feedback, RAG, preference learning, and RLHF-inspired approaches;

• Select candidate techniques to convert human feedback into reusable memory or operational guidance for agents;

• Implement local experiments with controlled examples and/or anonymized real cases;

• Create a mechanism for simulating and recording human feedback in the agent's execution loop;

• Define evaluation metrics such as success rate, error reduction, stability, latency, cost, and comparison with a baseline without adaptive memory;

• Integrate the selected approach into at least one pilot agent from the existing ecosystem;

• Gather and analyze real interactions with human feedback;

• Iterate the solution based on observed results;

• Package the solution as a library or component reusable by other agents;

• Produce technical documentation, usage examples, and a final report detailing results, limitations, and future steps.

⛳️ Requirements

• Education: Master's degree

• Fields of study: Computer Engineering, Information Systems

• Intermediate level Python skills

• Basic knowledge of LLMs, prompts, agents, and API model usage

• Familiarity with Git, code organization, and reading existing projects

• Ability to manipulate structured data, logs, JSON, and tables

• Basic SQL knowledge or willingness to learn quickly

• Capability to conduct experiments, define metrics, and compare approaches

• Strong written communication skills for producing technical documentation and reports

• Familiarity with development in real technical environments, including API usage, environment variables, authentication, logs, and service integration.

• Experience with agent frameworks such as smolagents, LangChain, LangGraph, CrewAI, AutoGen, or similar (preferred)

• Knowledge of RAG, embeddings, vector search, or semantic memory (preferred)

• Understanding of LLMs, agents, and pipelines evaluation with human-in-the-loop (preferred)

• Basic knowledge of reinforcement learning, RLHF, preference learning, or adaptive systems (preferred)

• Basic knowledge of AWS or another cloud platform (preferred)

• Familiarity with services such as S3, CloudWatch, IAM, Lambda, ECS/ECR, API Gateway, managed databases, Athena, Secrets Manager, Parameter Store, or Amazon Bedrock (preferred)

• Understanding of deployment, observability, logs, permissions, and security in cloud applications (preferred)

• Interest in AI applications within the legal domain.

🏝️ Benefits

• Opportunity to work in a dynamic and innovative environment

• Access to cutting-edge technology and resources

• Professional development and training opportunities

• Collaborative team culture and supportive work atmosphere