Projects

AI coding agent for the terminal, written in Rust. It reads the codebase, builds a plan, edits files, and runs commands — keeping the user in control of every consequential action.

• Full agentic loop (plan → edit → run → validate) in Rust
• LLM and terminal tooling integration with per-action approval
• Focus on safety and control: no destructive action without confirmation

llama.cpp runtime manager for running LLMs locally (Llama Manager). Ships runtime bundles with a versioned manifest; private source code with public releases.

• 100% local language model execution, no cloud dependency
• Distribution of llama.cpp runtime bundles with a version manifest
• Cross-platform model and runtime management

High-performance server written in Rust, focused on concurrency, low latency, and efficient resource usage.

• Asynchronous Rust architecture focused on throughput and latency
• Private project under active development

Multiplayer game server written in Rust — world simulation, networking, and real-time state persistence.

• Multiplayer server engine with real-time distributed state
• Private project under active development

High-performance infrastructure core written in Rust, foundation for other projects in the ecosystem.

• Reusable infrastructure components in Rust
• Private project under active development

Centralization and optimization of data for the Minas Gerais State Highway Department on Azure: multi-source ingestion, Data Lake storage, and reliable delivery to the business.

• Data Lake + SQL Data Warehouse + Databricks architecture on Azure
• Ingestion and distributed processing pipelines with Spark
• IaC with ARM Templates and CI/CD with Azure DevOps
• Security, encryption, and data privacy compliance

Data and AI initiatives in the financial domain: ETL/ELT with Spark, production OCR, RAG with vector stores, and fine-tuning for Q&A and classification.

• Production OCR (Tesseract, Azure Document Intelligence, Google Vision) with field validation
• RAG with embeddings, hybrid search, and anchored citations (FAISS/Pinecone/Chroma)
• Fine-tuning/LoRA for the financial domain; MLOps with AKS deployment
• Less rework through automatic normalization and fewer false positives in searches

Evolution of the continuous delivery culture: CI/CD pipelines, container orchestration in cloud and on-premises, and real-time observability.

• Optimized CI/CD pipelines with Azure DevOps
• Kubernetes and Docker in cloud and on-premises
• Logging and observability for real-time troubleshooting

Enterprise systems for the wholesale and retail market: .NET backend integrated with ERPs, RabbitMQ messaging, and Oracle/PostgreSQL databases.

• APIs and process automation integrated with ERPs
• Asynchronous messaging with RabbitMQ
• Front-end with Angular and React.js (PWA)