Lucas Gomide aka luzk

Build the platform and open-source runtime for deploying, observing and governing AI agents across CrewAI OSS, Enterprise and the Agent Control Plane.

• Designed and delivered an end-to-end AI policy system, from runtime enforcement in CrewAI OSS to policy authoring and deployment in the Agent Control Plane, backed by a proof of concept covering 21 policy cases.
• Rebuilt the Agent Control Plane metrics foundation on Cube, resolving slow dashboards and database out-of-memory failures.
• Established observability for critical platform services with Prometheus and Grafana, surfacing issues such as a 70% drop-off in an integration connection funnel and trace ingestion rejecting every sample.
• Delivered enterprise JIT SSO and team mapping for Entra ID, Okta, Auth0 and Keycloak, with setup guides for customer onboarding.
• Improved execution tracing across OSS and Enterprise, including human-in-the-loop support and fixes for incomplete or stuck traces.
• Shipped cost visibility for customers, with cost per execution and deployment-level spend alerts.

• Built Price Sheet Matching, which catches pricing errors on invoices so customers can issue credits or recover overcharges, recovering hundreds of thousands of dollars within a few iterations. Since products were not pre-registered, identified them using embeddings, vector similarity search, classification and AI agents.
• Built an API gateway that distributes invoice-processing requests across multiple data-extraction systems.
• Worked on OCR-based invoice detection and reading, and optimized the back-office workflow for processing invoices.
• Improved developer experience by introducing APM and error tracking and modernizing the codebase and testing tools.

• Supported cross-team architecture and design decisions, designing large-scale solutions for distributed systems.
• Evaluated existing processes to find and prioritize optimization opportunities.
• Drove AI advocacy and mentorship across engineering.

• Designed and led Mass Guidelines, a system processing a million patient guidelines daily, built on efficient data pipelines, Google Cloud services and distributed task execution to stay scalable and reliable as volume grew.
• Managed 5+ developers, growing their technical and soft skills through regular 1:1s and 180°/360° performance reviews.
• Partnered with Product from concept to delivery, leading solution proposals and their development.
• Championed measuring feature results against business metrics.
• Evolved an ecosystem of 10+ microservices.

• Modernized APIs and backend architecture to speed up feature development, improving scalability and instrumentation.
• Defined and implemented coding best practices and CI/CD.
• Organized hiring and interviewed for backend and frontend roles.

• Rebuilt the platform to support the company's move from e-commerce to B2B subscriptions in a few weeks.