Staff ML Engineer at MercadoLibre. I build ML systems for Display Ads — bidding, attribution, and GenAI at scale.
PhD in Optimization of Systems (UNICAMP, Brazil). Former mechatronics engineer. Still applying control theory to ML problems that others can't formulate.
At MercadoLibre, I lead ML systems for Display Ads. Not models in notebooks — systems that handle billions of impressions daily, with latency budgets measured in milliseconds. The work spans bidding optimization, pacing algorithms, attribution, and GenAI for content moderation.
Before MELI, I spent 6 years at UNICAMP (Brazil) doing a PhD in optimization, control theory, and neuroevolution. I built controllers for autonomous robots using genetic algorithms and potential fields. I still use that mindset — system dynamics, feedback loops, stability analysis — to design ML systems that don't collapse under load.
- Newsletter:Más Allá del Notebook — One story per week about ML in production. Spanish.
- Website:camilo-cf.github.io — Longer posts, technical deep dives.
Topics: ML in production, control theory for ML, the Senior-to-Staff transition, LATAM tech reality.
| Project | What it is | Stack |
|---|---|---|
| Robotarm-ABB140-AutoDrawing | Industrial robot draws arbitrary images from camera input | xBase, Robotics |
| llm-url_video-rag | RAG over YouTube videos and web content with LLMs | Python, LangChain |
| restricted-diet-cookbook-ai | AI cookbook for dietary restrictions with recipe generation | TypeScript, FastAPI, Next.js |
| rpi4-PID-fan-control | PID controller for Raspberry Pi 4 CPU temperature | Python, Control Theory |
Languages: Python, SQL, C (embedded), Bash
ML/DS: PyTorch, TensorFlow, Scikit-learn, XGBoost, Pandas, NumPy
ML Systems: Databricks, MLflow, Apache Spark, Airflow
MLOps: Docker, Git, CI/CD
Cloud: AWS, GCP
I don't list everything I touch. These are what I ship with.
PhD thesis: "Methodological proposal for optimization based on AI, data science, and Industry 4.0" (UNICAMP, 2020)
Google Scholar — 230+ citations. Most cited: "Control structure for a car-like robot using artificial neural networks and genetic algorithms" (Springer, 44 citations).
Research interests: Neuroevolution, control theory, optimization, system dynamics. I don't publish papers anymore — I ship systems. But the theoretical foundation still matters.
See what I'm working on now (or will be, once I create the page).
Not looking for a job. But always open to discuss ML systems, control theory, or LATAM tech.