Erick Gomes

Professor da Disciplinas de Redes Neurais Artificiais, Deep Learning e Algoritmos Genéticos. Professor da Disciplina de Generative AI and Advanced Nets.

As a Senior Lead, I am responsible for the architecture and operation of the data infrastructure, MLOps pipelines, and the entire machine learning model ecosystem. I lead strategic initiatives that combine traditional AI and Generative AI, ensuring reliability, scalability, and direct business impact. I conduct structured experiments with various state-of-the-art LLMs (GPT, Claude, Gemini, Llama, DeepSeek, BERT-based models), comparing performance, costs, and suitability for real-world use cases. I develop complete GenAI solutions, from prototyping to production integration, using frameworks such as LangChain, LangGraph, LangFlow, LangFuse, n8n, Flowise, and optimized RAG architectures. I lead projects involving: Intelligent information extraction (IE) and process automation using generative models. Advanced OCR and parsing solutions with LLMs, combining computer vision, RAG, and structured prompts. Building automated agents and flows with LLMs + data pipelines, integrating internal and external systems. Voice projects (STT/TTS) with cutting-edge technologies such as Deepgram, ElevenLabs, Groq, and Whisper, creating multimodal assistants and flows. Scalable MLOps architectures for ML and GenAI models, ensuring monitoring, versioning, traceability, and governance.

Responsible for the data ingestion pipeline of the MLOPs pipeline using AWS (S3, Glue, Athena, EC2). -> Building python codes and refactoring for integration with AWS via SDK. Responsible for the MLOps pipeline using MLFLOW, DVC, Github Actions, Evidentlyai, EC2, Airflow, Prometheus, Grafana and Terraform. Responsible for building machine learning models, deployment and maintenance. -> Building a propensity model for the collection squad. -> Building a sales forecast model. Responsible for building and maintaining data engineering pipelines using AWS(Glue, Lambda, Redshift). -> Consumption of transactional database (OLTP) for analytical environment (OLAP). -> Consumption of API data (REST and GraphQL) for analytical environment. Building Chatbots using generative AI, langchain, Vertex AI (Agent Builder). -> Using langchain to integrate with various GenAI models. -> Using LangChain to build Retrieval Augmented Generation (RAG). -> Using LangChain, CrewAI, AutoGen and Langflow to build AI Agents. -> Building systems with multiple AI agents. -> Monitoring Agents. Responsible for managing the team's Backlog and leading the scrum ceremonies.

Responsible for the Course: Model Production and MLOPs Course Outline: MLOps concepts: integration between machine learning and DevOps; Lifecycle of a machine learning model in production; Data and model versioning; Model monitoring: performance, drift, and metrics in production; Machine learning pipeline automation: CI/CD for models; MLOps tools: MLflow, Kubeflow, TFX Model deployment: APIs, containers (Docker), and orchestration with Kubernetes; Scalability and optimization of models in production and FinOps; Security and compliance practices for models in production. Responsible for the content production of the Course: Database Course Outline: Modeling of databases SGBDs SQL: DML/DDL SQL: DQL SQL: Joins

Developing my research in the area of ​​machine learning applied to Physics. I have been using good data science practices to propose new solutions in the analysis of large databases. Main Results: 1. Use of regression algorithms to build predictive models in order to determine the formation energy of materials and, with this, establish a set of materials with thermodynamic stability. 2. Use of classification algorithms to build a predictive model in order to separate metallic and insulating materials. 3. Construction of a regression model to predict various properties of insulating materials. Tools: Python, Scikit-learn, Pandas, Numpy, Seaborn, C2DB-database.

As a tutor for the Machine Learning Applied to Physics course, I played a central role in exploring and applying the fundamental principles of machine learning to solve specific challenges within the scope of theoretical and experimental physics. My technical contribution involved guiding students in the implementation of supervised and unsupervised learning algorithms, such as regression, classification, neural networks, and clustering methods, to analyze data sets from diverse sources. Using languages ​​and libraries such as Python, TensorFlow, and scikit-learn, I explored data preprocessing, feature selection, and model optimization methods to extract accurate insights and reliable predictions. I was responsible for the weekly guidance of the postgraduate course and assisting in the completion of the course activities.

Responsible for developing dashboards using Power BI. Responsible for developing process automation and building ETL (Extraction, Transformation and Load) pipelines and creating databases. Development of churn models using machine learning algorithms for financial institution clients.

- Analyze Data. - Develop proof of concepts on the analyzed data. - Structure financial information generating business insights. - Study market trends focused on Open Finance. - Issue reports (for the requesting business area). - Build a model to predict default. - Build a model to detect Fraud in Online Payments. - Use of techniques for data balancing. - Evaluation metrics for classification models (AUC, Precision, Recall, F1-Score, AUPRC). - Model Deployment. Responsible for building the sandbox with regulated Open Finance data. Responsible for executing the ETL pipeline to load simulated data into the database. Responsible for creating the investment copilot using LLMs and open data from Open Finance.

Main Results: 1. Construction of a regression model to forecast sales of a chain of stores.

I worked with computational simulation of materials. During the project I used C, Shell Script, developed skills with Linux systems, and had experience in using high-performance computing environments. Responsible for developing computational parallelism routines for massive data processing using C. Responsible for developing shell scripts (linux) for task automation. Responsible for building data visualization using different visualization software.

Principais Disciplinas: - Aprendizado de Máquina. - Mineração de Dados. - Mecânica Quântica.

Principais Disciplinas: Big Data Pipelines Big Data Cloud Platforms Big Data Storage Structures Aprendizado por reforço Sistema de recomendação Redes Neurais e Deep Learning NLP Computational Vision Generative AI Arquitetura de Feature Store Conteiners MLOps

Principais Disciplinas: - Cálculo I, II, III e IV. - Física I, II, III e IV. - Álgebra Linear e Geometria Analítica. - Algoritmos. - Estrutura de Dados. - Programação Orientada a Objeto. - Física-Matemática I e II. - Mecânica Quântica I e II. - Eletromagnestimo I e II. - Física Estatística. - Métodos Numéricos I e II. - Computação de Alto Desempenho I e II.

Principais Disciplinas: - Eletrônica Analógica - Eletrônica Digital - Controlador Lógico Programável (CLP). - Programação em Assembly. - Programação em C.