RAG Chatbot
Cloud RAG for the branch network & contact center. Evals scaled from <100 to >1,000, eval time cut 10×. V1.0 in production at 20% adoption, V2 shipping next.
AI Engineer and Technical Product Manager. Eight years bridging technical execution and business strategy across AWS, Azure & GCP.
One profile that has held every technical role in the data stack, and now builds AI products that ship.
From RAG product management to hands-on AI engineering: building and shipping AI systems for one of Mexico's largest banks, end to end.
Cloud RAG for the branch network & contact center. Evals scaled from <100 to >1,000, eval time cut 10×. V1.0 in production at 20% adoption, V2 shipping next.
Taught 300+ employees across 6 bank areas on agent harnesses. Built internal AI tooling now used by 50+ engineers.
Agent over a knowledge base with two tools: structured SQL + unstructured OCR RAG. Being deployed to production AWS as KB v0.1.
Structured fraud-analysis agent
Santander Mexico
A prototype Santander agent harness built to break a budget constraint: only 10% of the Mexico org had a ChatGPT license.
The project influenced the global organization to make AI available to every Santander Mexico employee, today delivered through Copilot Chat.
CI/CD (MLOps) for an international bank across Snowflake ML, VertexAI, Sagemaker, AzureML & Databricks. Legacy proprietary autoML migrated to open-source equivalents.
✓ V1.0 delivered to productionFull data value-chain assessment for an international insurer: OnPrem DBs, DWH, cloud migration & BI. Diagnosed bottlenecks from poor modelling & missing analytics engineering.
✓ Prioritized roadmap deliveredUnified brick-&-mortar + e-commerce KPIs for an international retailer. ETLs built from scratch in PySpark; semantic layer & pipelines still in use today.
✓ Still running in productionPrior: Software Engineer · Development Bank (2020–21) · Data Scientist · Central Tax Administration (2018–20) · Data Engineer · Boutique consulting (2017)
Responses are generated using AI and may contain mistakes.