Luca Pinheiro Gomes

Senior AI Architect at Inteligência Itaú (Itaú Unibanco), working on modernization initiatives across cards and retail (PF) banking products. I design and build agentic systems and agentic platforms that power conversational and transactional experiences across multiple channels. I act as a technical reference in agentic architecture, contributing directly to the patterns, frameworks, and design principles that shape how AI agents are built and operated at the bank. I was one of the main contributors to the reference agentic architecture adopted across Itaú, introducing patterns, best practices, and innovation now leveraged by multiple teams. My work spans the full lifecycle of agentic solutions, from high-level architecture and multi-agent orchestration design to production concerns such as scalability, observability, safety, latency, and integration with core banking systems. I also actively contribute to the internal Itaú technical community, sharing knowledge, mentoring engineers, and driving adoption of agentic patterns across the organization. Tech Stack: Python, FastAPI, LangChain, LangGraph, LLMs, Multi-Agent Systems, MCP, RAG, Vector DB, Microservices, Distributed Architecture, AWS, Azure, Kubernetes, CI/CD, Observability

AI Architect at XP Inc., responsible for designing and building the scaffold agent runtime of XP's first Generative AI platform, the foundation that enables business areas across the company to build and deploy agentic solutions in a standardized, scalable, and secure way. I was working on the architecture and implementation of the platform's main functionalities, including support for multi-agent patterns, MCP (Model Context Protocol) server integration, standardized tool creation and registration, and observability for agentic workflows. The platform was designed from the ground up to abstract complexity from product teams, allowing them to focus on business logic while relying on a robust, production-grade runtime. Beyond the technical foundation, I worked closely with multiple business areas to onboard their first GenAI use cases, providing architectural guidance on agent design, integration patterns, performance, and user experience. Tech Stack: Python, FastAPI, LangChain, LLMs, Azure OpenAI, Azure, AKS, Microservices, Distributed Architecture, Multi-Agent Systems, MCP, RAG, Vector DB, CI/CD

Software Engineer at XP Investimentos, working within the Open Finance ecosystem across engineering, business, and governance teams, contributing to highly critical financial systems with strong focus on scalability, reliability, and regulatory compliance. I am the principal engineer in the team driving innovation, leading initiatives focused on automation and applied AI to improve incident management, observability, and operational efficiency. Guiding the team on how to accelerate their work using GitHub Copilot, GitHub MCP, and AI features that speed up development. Key contributions include the development and sustainment of more than 20 Open Finance microservices, supporting critical domains such as investments, personal and corporate resources and financial / investment products. I also participated in the development of XP’s Open Finance Credit Portability API. I actively work on production reliability and incident management, participating in critical rooms and performing end-to-end troubleshooting, from infrastructure analysis to code-level fixes. I designed and implemented a data engineering solution to track regulatory incidents opened by BACEN and other financial institutions, built an executive Power BI dashboard used by leadership, and automated the creation of Azure DevOps work items from BACEN Service Desk incidents, eliminating manual operational effort. Additionally, I participated a FinOps-focused optimization of Azure AKS workloads, improving resource efficiency through right-sizing of clusters, pods, and compute, validated with Dynatrace and Grafana. I also provide architectural guidance and designed the Personal and Corporate Income Tax solution planned for 2026–2027. Tech Stack: .NET, C#, Python, Azure, AKS, Design System, Microservices, Distributed Architecture, Azure DevOps, CI/CD, Dynatrace, Grafana, Power BI, Open Finance, Data Engineering, Automation, AI, Copilot, MCP

Role Highlights 1. Deployment Risk Prevention & Automation • Developed a robust GitHub Actions step in TypeScript to validate configuration and ensure production readiness for container-based applications across OpenShift, AKS (Azure Kubernetes Service), and EKS (Amazon Elastic Kubernetes Service). • Partnered with DevOps, Governance, and Security teams to implement compliance rules, enforce best practices, and automatically block non-conformant deployments. • Designed logic for automatic deployment blocking and real-time notifications to responsible teams, reducing human error and improving operational resilience. • Impact: • Over 600 misconfigured deployments were blocked in 2025, preventing potential production incidents. • Achieved zero critical outages related to misconfigurations post-implementation. • Significantly improved governance, reliability, and cost-efficiency in CI/CD pipelines. Tech Stack: TypeScript, Node.js, GitHub Actions, Kubernetes (AKS/EKS), OpenShift, Azure, AWS, Docker. ⸻ 2. GenAI Chatbot for Incident Management • Designed and deployed an internal GenAI chatbot to improve incident resolution across multiple departments, leveraging Large Language Models (LLMs) and autonomous agent orchestration. • Implemented RAG (Retrieval-Augmented Generation) pipeline using Azure OpenAI, vector databases, and contextual search for precise and domain-specific answers. • Built a backend with FastAPI and Python, integrated with Postgres for structured data, and used LangChain for orchestration and agent logic. • Ensured seamless integration with internal systems to accelerate incident triage and resolution, reducing response times and improving overall operational efficiency. Tech Stack: FastAPI, Python, Azure OpenAI (LLMs & embeddings), Vector DB, Postgres, Java Spring.

Main Contributions During at Santander During at Santander, I was responsible for developing the area's main market share dashboard, using a database with more than 20 million rows. This solution allowed detailed analyses of Santander's performance and that of its competitors, providing strategic support for decision-making. I restructured six ETL processes in Python, reducing the execution time from 1h20 to just 6min25s, using advanced data science techniques, such as vectorization with Pandas and optimizations with Spark. This improvement freed up computational resources, mitigating CPU/VM instance costs and increasing the efficiency of strategic analyses. These optimized ETLs now serve more than 800 specialists and around 23,000 stores throughout Brazil, consolidating more than 60 data sources into just six processes. I applied advanced data engineering practices and configured CD flows to keep the portal always updated and accessible via mobile app. I participated in the area's process of migrating approximately 180 tables from the on-premise environment to the cloud in Databricks, using PySpark, converting the data to Parquet format and dealing with a big data environment, ensuring scalability and performance. I was a reference in development and innovation, I designed and designed a multi-agent system based on Large Language Models (LLM) ready to be developed and serve +800 experts throughout Brazil, offering insights based on GenAI and automating complex analytical tasks. My main skills include: Data Science & Data Engineering: Python (Pandas, PySpark, NumPy, Plotly, SKLearn), ETL, Machine Learning Data Management: SQL, Databricks, Cloud Migration Data Visualization: Power BI (Dashboards, DAX) GenAI (Azure AI Tools, Langchain, CrewAI, OpenAI)

Software Engineer Responsible for developing and maintaining scalable microservices with a strong focus on clean code, automation, and cloud-native practices. Actively work within a multidisciplinary team involving governance, DevOps, quality, and cybersecurity to ensure delivery excellence and proactive issue resolution. Tech Stack: • Java (Spring, Groovy), Node.js (NestJs) • CI/CD pipelines • DevOps practices • Azure & AWS Cloud, OpenShift • Kubernetes (K8s) • Git, Docker AI Engineer Collaborating on the design and deployment of an internal GenAI chatbot solution aimed at improving incident handling across internal departments. The solution leverages LLMs, autonomous agents, orchestration logic, and vector search with RAG (Retrieval-Augmented Generation) techniques for contextual responses. Tech Stack: • FastAPI, Python • Azure OpenAI (LLMs & embeddings) • Vector DB (RAG indexing) • Postgres • Java Spring (Arsenal) • LangChain

As Head of Backend Development for the startup, I was responsible for architecting the entire API, database, and infrastructure of a social networking app designed to help users find the best nightlife events and interact with friends seamlessly. The backend was developed using Node.js with TypeScript, following SOLID principles, Design Patterns, and a Clean Architecture approach focused on microservices. The infrastructure was entirely based on AWS services, leveraging Route 53 for DNS, CloudFront for content delivery, Cognito for secure authentication, Rekognition for user photo validation, Lambda for serverless functions, API Gateway for API management, and S3 for file storage. The MongoDB NoSQL database was hosted on a dedicated cluster to handle big data and high traffic. I led a team of 10 developers, managing all backend development stages from architecture to deployment. I was responsible for designing the complete system architecture, implementing DevOps practices, and creating environments for testing and homologation, including APK and TestFlight distribution for seamless testing on mobile platforms. Additionally, I managed the app’s design system, ensuring a consistent and intuitive user experience across iOS and Android. I oversaw the app’s publication on the App Store and Google Play Store, handling technical and administrative compliance for a smooth launch. My work combined serverless architecture, microservices, and cloud-native technologies, resulting in a backend that is scalable, maintainable, and prepared to support real-time interactions and high user traffic. My leadership kept the development team aligned with business goals, delivering a high-quality product on time and within scope.

As a software developer and responsible for infrastructure, I am responsible for developing a full-stack application for the teachers academy at Instituto Mauá de Tecnologia. My role is as a technical leader of the application, focusing on the back-end. The system is about managing events, lectures, registrations, recording attendance and downloading a certificate! Furthermore, the project has an administration area that allows the administrator to have full control over events and teacher certificates. I built the application from scratch using good development practices and clean architecture, I used the ExpressJS framework with NodeJS Typescript, Postgresql relational database with the Prisma ORM. I used cloud services, such as Object Storage from Oracle Cloud, to store the certificates, and I instantiated a VM (Oracle Cloud) to host the server. Additionally, I use AWS services such as EC2 and also Heroku to secure the server and S3 to save the attendance certificates for each event attended by a teacher, thus ensuring the stability and security of the system. Como desenvolvedor de software e responsável pela infraestrutura, lidero o desenvolvimento de uma aplicação full-stack para a academia de professores do Instituto Mauá de Tecnologia, com foco no back-end. O sistema gerencia eventos, palestras, inscrições, presenças e downloads de certificados, além de contar com uma área de administração para controle total de eventos e certificados de professores. Construí a aplicação do zero utilizando boas práticas de desenvolvimento e clean architecture, com ExpressJS e Typescript no NodeJS, e PostgreSQL com a ORM Prisma. Utilizo serviços em nuvem, como Object Storage da Oracle Cloud para armazenar certificados e uma VM na Oracle Cloud para hospedar o servidor. Também uso serviços da AWS, como EC2 e S3, além do Heroku, garantindo estabilidade e segurança ao sistema.

As a Back-end developer, I am involved in the development of MauaFood, a platform for scheduling and ordering college meals and products. We use pure Python and clean architecture, with Lambda functions on AWS for an efficient and scalable serverless application. Additionally, we implemented websockets for real-time communication about order updates, ensuring an even more fluid and interactive experience for users. Our goal is to provide an impeccable experience, simplifying the food process on campus.

My role during my time at the company was to read, analyze documentation and check if it was possible to integrate. After that, I created adapters in the company's repository with the aim of integrating a clinic's external API with our API. I used Python to make the requests and create the methods to use in bot production. I used Pytest to unit test all developed methods. After that, I tested and executed the request in postman! Valuable experience where I learned a lot in a short period of time.

Bacharelado de Ciência da Computação no Instituto Mauá de Tecnologia