Ernesto Vizcaíno
AI / Product Engineer
I've been building software since I was 13. Since then, I've taken products from idea to thousands of users, built revenue generating SaaS and fintech systems, and applied machine learning to millions of astronomical objects.
Pressto copy my emailCopied
Experience
I work across engineering and product, taking ideas from prototype to real users and production.
AI Engineer / Founding Team atFinanciamiento Inteligente / Xignus
Built a credit pre approval system that supported more than $4M MXN in SMB financing during its first month, alongside production AI agents and financial data products. Rejoined in 2026 to build CONTPAQi integrations and ETL pipelines that transform accounting data into automated Excel reports and reconciliation workflows.
Technical Cofounder atOliver AI
Led engineering across a fintech platform and production AI workflows for loan origination, portfolio analysis, risk review, compliance and financial document processing.
Founder / Engineer atOliver POS / ERP
Built and launched a mobile POS that reached 10,000+ downloads and 3,000 active users within three months, then expanded it into a complete web based ERP.
Awards
1st Place, Xólotl Hackathon
Phase 3, Advanced Track · CUDI / LAMOD UNAM
Rubin/LSST Astronomical Time Series Classification
Won the advanced track after building an uncertainty aware classification and scientific prioritization pipeline for Rubin/LSST like astronomical time series data.
- 5.1M+ astronomical objects
- 10 morphologies
- 6,000 evaluation light curves
- 600 completely unseen observing cadences
- 68.47% balanced accuracy
- 68.28% macro F1
View official announcementAward includes attendance at CARLA 2026 in Córdoba, Argentina 🇦🇷.
Selected work
Selected products, systems and research I've built.
- Rubin/LSST Time Series Classifier1st PlaceMachine Learning · Astronomy1st Place · 5.1M+ objects · 68.47% balanced accuracyUncertainty aware astronomical classification and scientific prioritization pipeline built for Rubin/LSST like time series data.
- Gaia / OGLE Variable Star ClassifierResearch · Machine Learning491K Gaia sources · 11 classes · 0.9847 weighted F1Cross survey variable star classification using Gaia light curves, period, Fourier, color and catalog features, validated against OGLE labels.
- Oliver POS / ERPProduct · SaaS10K+ downloads · 3K active users in 3 monthsMobile POS for small businesses that later expanded into a full web based ERP.
Research
Gaia / OGLE Variable Star Classification
Universidad Autónoma de San Luis Potosí
Built a machine learning pipeline for periodic variable star classification, processing 491,073 Gaia sources and engineering 91 time series, Fourier, color and catalog features. A class weighted XGBoost model reached 0.9495 macro F1 on held out data and 0.9847 weighted F1 against mapped OGLE labels.
Stack
Contact
I'm open to product engineering opportunities, ambitious startup teams and selected collaborations.