I'm an MLOps / Machine Learning Engineer at Capgemini, with a background in Engineering Physics. I design and run end-to-end production ML systems on AWS and GCP across banking, retail and IoT, from fraud detection and demand forecasting to constraint optimization and generative AI.
What interests me is what happens after the model works in a notebook. Most of the value in machine learning is lost in the gap between a working model and a system a business can rely on, and that gap is where I work: turning experiments into pipelines that train, deploy, monitor and retrain themselves, with the governance and reproducibility production demands.
I own the full lifecycle: data pipelines and feature stores, CI/CD and Infrastructure as Code, containerized deployment, model monitoring and drift detection, automated retraining, and both batch and real-time inference.
My current side project is coffee-mlops: a domain-reusable MLOps framework that runs end to end on a laptop, bootstrapped on coffee from the farm to Mexico City's coffee shops.
Outside of ML, I'm into video game development. I'm building two games in Unity 6 and C#: a co-op multiplayer action-RPG roguelike and a turn-based tactical card RPG.
Machine Learning Engineer · Capgemini
Oct 2026 – Present · Mexico City · Hybrid
Machine Learning Engineer · Paco App
Nov 2025 – Jun 2026 · Mexico City · Retail
- Built an end-to-end workforce-scheduling ML system on SageMaker Pipelines + Step Functions: K-means store segmentation → XGBoost sales forecasting → OR-Tools shift optimization. Cut overtime by 30%
- Parallelized clustering (10 h → 2 h), training (8 h → 2 h) and optimization inference across all stores (3 h → 30 min) for an automated weekly cadence
- Full AWS stack as code with CloudFormation via Azure DevOps; MLflow tracking with drift-triggered retraining; batch + real-time inference through a Dockerized FastAPI service
- Architected a multi-agent GPT-4o chatbot using MCP to query MariaDB and the workforce app in natural language
MLOps Engineer · Gentera
Mar 2024 – Sep 2025 · Mexico City · Banking
- Designed and maintained batch MLOps pipelines for fraud detection on GCP with Vertex AI + Kubeflow, orchestrated by Airflow on Cloud Composer
- CI/CD for automated retraining, validation and deployment; drift monitoring with BigQuery and Evidently AI; unit tests for DAGs and Vertex Pipelines
Data Scientist · ECOSAT
May 2023 – Mar 2024 · Chihuahua · IoT
- First data scientist at the company; set up the data science function and a company-wide data-driven strategy
- SVM anomaly detector, churn prediction pipeline with a Flask push-notification API, ETL automation with Dagster + dbt, and Power BI dashboards
Scale AI · Web Scraping Parsing Engineer, previously Technical QA & QA Team Lead
Aug 2021 – May 2023 · Remote
- Built scraping and parsing scripts (Node.js, Puppeteer) for 30+ websites per week, processing data on AWS Lambda into AI training datasets; cut validation time by 60% with automated testing
- Led and trained a QA team, and was promoted for 3× productivity benchmarks
| Area | Tools |
|---|---|
| AWS | SageMaker Pipelines · Step Functions · Lambda · Glue · S3 · ECR · EventBridge · API Gateway · CloudFormation |
| GCP | Vertex AI · BigQuery · Cloud Composer · Cloud Run · Pub/Sub |
| MLOps | MLflow · Kubeflow · Apache Airflow · Dagster · Evidently AI · DVC · Pandera |
| CI/CD & Infra | Azure DevOps · Jenkins · Docker · Kubernetes · Infrastructure as Code · pytest |
| ML & Optimization | XGBoost · LightGBM · scikit-learn · PyTorch · TensorFlow · OR-Tools · FAISS · Optuna |
| Generative AI | GPT-4o · Multi-agent systems · Model Context Protocol (MCP) · LangGraph · RAG with Qdrant · Ollama |
| Data & BI | Pandas · Polars · DuckDB · Apache Spark · Databricks · dbt · Azure Data Factory · SQL Server · MySQL · MariaDB · Power BI |
| APIs & Apps | FastAPI · Flask · Streamlit |
| Languages | Python (uv) · SQL · JavaScript (Node.js) · Bash |
| Game Dev | Unity 6 · C# · URP · UI Toolkit · Netcode for GameObjects |
| Exploring | Reinforcement Learning |
| Project | What it does | Stack |
|---|---|---|
| Coffee MLOps In development |
Domain-reusable MLOps framework that runs fully local, built on coffee from the farm to the cup: 25 sources held to data contracts, nine models promoted only past a paired-bootstrap gate, drift monitoring, and a LangGraph agent over SQL, the prediction API and Qdrant, served through FastAPI, an MCP server and a Streamlit explorer | Polars · DuckDB · LightGBM · MLflow · Evidently · LangGraph · Qdrant · FastAPI · MCP |
| Unity roguelike videogame In development |
3D co-op action-RPG roguelike for 1–4 players: create a character, pick a class and fight through 12 increasingly harder areas per run, with JRPG-style progression, four independently charging skills and bosses that scale with party size | Unity 6 · C# · URP · Netcode for GameObjects |
| Unity tactical card RPG In development |
Turn-based tactics on a 20×20 board crossed with a Yu-Gi-Oh!-style card game: 179 cards across six elemental factions, a reaction chain, fusion and ritual summons, and an AI personality per faction, on a deterministic C# rules engine tested with golden AI-vs-AI replays | Unity 6 · C# · URP · UI Toolkit · NUnit |
| Azure Predictive Maintenance | Production-ready template that predicts machine failures 24h ahead from sensor telemetry: rolling-window features, config-driven runs, and a FastAPI serving API | Python · scikit-learn · Hydra · Docker · FastAPI |
| Steam Game Recommender | Two-tower retrieval model with a FAISS index for fast similarity search, served through FastAPI with a Streamlit dashboard | PyTorch · FAISS · FastAPI · DVC · Docker |
| Music Generation | Transformer that generates MIDI tracks by instrument, with hyperparameter tuning and a web UI for generation | PyTorch · CUDA · Optuna · FastAPI |
| Binary Classification Pipeline | Modular, tested pipeline for tabular binary classification that writes models, metrics and plots as artifacts | Python · scikit-learn · pytest |
Also from DeepLearning.AI (2023): Neural Networks and Deep Learning · Structuring Machine Learning Projects
Engineering Physics, Universidad Autónoma de Chihuahua (2015–2019)
Publication: Construyendo el universo con herramientas. Revista Ciencias, UNAM (No. 131–132, 2019). An essay on how we perceive our environment through the tools we build, and how those tools have evolved over time.