I build AI/ML systems across LLMs, multi-agent systems, RAG, NLP, multimodal AI, and backend infrastructure.
My work spans both research and product engineering, from multilingual NLP research with the ALGORITMI Centre at the University of Minho and SemEval 2026 to building AI-powered products as a co-founder. I focus on turning research ideas into reliable, scalable systems, with experience in conversational AI, fraud prevention, speech processing, Text-to-SQL, and distributed backend systems.
- 💼 Co-Founder & AI/ML Engineer @ Remy.AI — AI-powered fraud prevention through WhatsApp & Telegram.
- 🚀 Co-Founder & Software/ML Engineer @ Houmy — Building an Uber-like platform with Rust-based geospatial matchmaking and distributed backend systems.
- 🔬 Research: SemEval 2026, ALGORITMI Centre / University of Minho, RAIA & LABIC @ ICMC-USP.
- 🎓 Education: B.Sc. Data Science at University of São Paulo (USP), with an emphasis on Artificial Intelligence.
- 🏆 SemEval 2026: Co-authored a published system description paper; ranked 7th/16 among complete multilingual systems.
- 🧠 Focus: LLMs, Multi-Agent Systems, RAG, NLP, Multimodal AI, ML Systems, and AI infrastructure.
- SemEval-2026 Task 9 — Developed a selective multilingual and multitask specialization approach for online polarization detection in collaboration with the ALGORITMI Centre, University of Minho. Ranked 7th/16 among complete multilingual systems.
- TextToInsight — Research and engineering on multi-agent Text-to-SQL systems using LangGraph, specialized agents, iterative self-correction, and Human-in-the-Loop interaction.
- Large Audio Language Models for Emotion Recognition — Comparative evaluation of Audio Flamingo 3, Qwen2.5-Omni, and WhisperX + Mistral-7B across MELD and IEMOCAP, studying prompting, accuracy, latency, VRAM, and computational trade-offs.
- Remy.AI — AI-powered fraud prevention chatbot for WhatsApp and Telegram, combining LLMs, RAG, NLP, and lightweight ML routing. Validated through 300+ real-world interactions.
- Houmy — Uber-like service platform with a low-latency provider-client matchmaking system using Rust, Redis, H3 geospatial indexing, PostgreSQL, and asynchronous event-driven pipelines.
- TextToInsight — Open-source multi-agent Text-to-SQL system using LangGraph, with Planner, Code, Executor, and Critic agents, iterative self-correction, Human-in-the-Loop interaction, and automated visualization.
- ML-TutorBot — Multilingual AI tutor using RAG, modular agents, Chroma vector search, code execution, and language translation.
- Unfaker — Open-source Chrome extension and FastAPI backend for detecting potentially misleading social media content using NLP/ML models.
- ML Model Factory — End-to-end MLOps pipeline for automated data preprocessing, training, hyperparameter optimization, experiment tracking, and model versioning.
Programming:
Python, C/C++, Rust, TypeScript, JavaScript, SQL
AI/ML:
LLMs, RAG, Multi-Agent Systems, NLP, Multimodal AI, Machine Learning, PyTorch, TensorFlow, Scikit-learn, Transformers, Fine-tuning
AI Frameworks:
LangGraph, LangChain, Hugging Face, vLLM, Unsloth
Data & Infrastructure:
PostgreSQL, Redis, ChromaDB, Docker, RabbitMQ, H3, REST APIs, FastAPI
Cloud & MLOps:
AWS, Azure, MLflow, CI/CD, Experiment Tracking, Model Evaluation
