Consulting · MLOps, Data and Cloud
I design and industrialize data and MLOps platforms: CI/CD for models, pipelines that hold real volume, and a team of yours able to run it without me. Squad Lead of MLOps Engineering at BCP and 5+ years in Data Science, DevOps and Machine Learning.
Section 01 · Services
Four fronts. In all of them what I hand over is code in your repository and a team able to maintain it — not a slide deck.
Design and implementation of the model's full path: from experiment to automated deployment, with registry, versioning and monitoring. Cloud-native architecture on Azure, Databricks and Kubernetes.
You getA pipeline that deploys on its own, with its model registry and its alerts switched on.
Scalable data architectures with Azure Data Factory, Databricks and Apache Spark. Ingestion, quality and real-time processing when needed — and batch when that is the sensible answer.
You getVersioned pipelines, with tests, and data quality measured rather than assumed.
Development and optimization of models for a concrete use case, plus the visualization layer that makes someone actually decide with them: dashboards, KPIs and tracking.
You getA model measured against a baseline, and a dashboard that gets used.
Digital transformation strategy, technology selection and training for the in-house team. For when the problem is direction rather than code, or when the team already exists and needs tuning.
You getA prioritized plan, written best practices, and your people able to carry on alone.
Section 02 · Results
Real projects, with the impact measured where it was measured.
Logistics analytics product that contributed to a $15M profit increase in 2022. Full MLOps implementation with Azure, Kubernetes and MLflow for ML model management.
Machine Learning system to improve phone campaign contactability, resulting in a 20% increase in new customer acquisition at Interbank.
Intelligent post-sales chatbot built with Google Dialogflow and Flask. Winner of first place at Interbank's Innovatón 2020, significantly improving customer service.
Development and audit of 120 Power BI dashboards with 90% usage, significantly improving decision-making tools at QROMA.
Optimization of lead distribution algorithm for acquisition campaigns, reducing processing time from 4 hours to 10 seconds.
Every figure comes from the project where it was recorded; the organization and year sit next to the number.
Section 03 · How I work
The same order on every engagement. Each step closes with something that is yours, even if the work ends there.
I review your stack, your data and how your team works. You walk away with a map of what exists and a prioritized plan — yours even if we stop there.
One end-to-end case in production: data, model, deployment and monitoring. Small on purpose, so it can be seen working soon rather than in a year.
Whatever worked in the pilot becomes a template: Asset Bundles, GitHub Actions, governance and catalog. The second model costs a fraction of the first.
Documentation, training and support until your team runs it without me. If you need me forever, I did it wrong.
Everything is code, commits are semantic, and we trust the Delta Log.
Section 04 · Contact
Write me two lines about the problem and I will reply with what I would do and where I would start. No strings attached.
