ruancomelli

Developed LLM-powered systems for automated code review and bug fixing, deployed in Linux environments. Re-architected an event-driven GitHub/GitLab bot infrastructure supporting natural language comment commands. Implemented automated PR titles and summaries; Mermaid diagram visualization; and security-focused reviews — all widely adopted and loved by users. Built evaluation datasets and scripts for AI-generated patches, using LangSmith for testing and observability. Improved CLI, GitHub Actions, and IDE extensions to streamline workflows and increase adoption. Led codebase modernization by replacing Black and PyLint with Ruff, making linting and formatting near-instantaneous; upgraded Python/runtime dependencies for a smoother dev experience. Enhanced our static analysis and refactoring engine, adding linting and refactoring rules.

Led a team of 4 engineers in developing a proof-of-concept for automated PDF data extraction, successfully delivering a solution that was adopted by the client. Architected and implemented a custom OCR solution that outperformed state-of-the-art commercial tools, including Google Vision, by combining Tesseract with modern open-source models. Designed and implemented serverless document processing pipelines using infrastructure-as-code.