René Núñez

Data & AI Engineer in the San Diego–Tijuana area, with a background in behavioral neuroscience research, municipal government, and production statistical systems. Outside work I build probabilistic forecasting models for MLB, college football, the NFL, and the NHL, publish their numbers before the games start, and grade them against the closing line in public. The betting market is the benchmark because it is the sharpest public probability estimate available, not because the goal is a betting product.

ResumeResume (PDF)

GitHubgithub.com/renenunezg

LinkedInlinkedin.com/in/renenunezg

Twitter@nunezanalytics

Emailrenenunezgalaviz@gmail.com

Looking for

Data scientist, analytics engineer, and applied statistics or forecasting roles where calibrated probabilities and honest evaluation matter more than a point estimate. San Diego, Tijuana, or remote.

Strengths

Probabilistic modeling

Hierarchical Bayesian models fit with NUTS (PyMC, NumPyro, JAX), Monte Carlo simulation, and parameter uncertainty carried through to the published number.

Forecast evaluation

Calibration, Brier and log-loss scoring, walk-forward validation on untouched holdouts, and benchmarking against efficient markets, with the record kept in public.

Data engineering

Python, SQL and Postgres, ETL and scheduled pipelines on GitHub Actions and Supabase, with write-gated production boundaries.

Reporting

SQL reporting infrastructure, KPI dashboards in Power BI and Tableau, and the Next.js and TypeScript frontend that serves these models.

Experience

Data & AI Engineer · Group U

2026 – present

Account-level forecasts covering about 48,000 calls a week, built from historical trends, seasonality, and holidays to guide staffing decisions; headcount targets from a model of missed-call rate against staffing, plus the hiring scenarios behind staffing decisions; root-cause analyses of forecast misses; AI call-quality scoring with AssemblyAI and Claude, measured for error, bias, and agreement against supervisor grades; and automated executive KPI reporting on PostgreSQL and Azure SQL.

Data Science Research Volunteer · Salk Institute for Biological Studies

Jul 2025 – Mar 2026

Built the labeled dataset behind a SLEAP pose-estimation model from tube-test dominance video, resolved systematic labeling inconsistencies across cohorts, and cross-validated the model to decide which displacement classifications were reliable enough for downstream analysis.

Data Analyst · Tijuana City Council

Jan – Jul 2025

Beneficiary lookup dashboard deployed in an air-gapped environment for elected officials and staff; query optimization that cut lookup time 22% during live sessions; ETL automation that removed 30% of manual data handling.

Data Analyst · Executive Offices

2022 – 2024

SQL reporting infrastructure for billing, occupancy, and fiscal activity of Mexican companies entering the U.S. market; resolved duplicate records, manual-entry errors, and malformed dates, cutting the error rate 26%.

B.S. in Computer Science, CETYS Universidad (2025), with the EGEL-CENEVAL Sobresaliente distinction in Intelligent Computing, and an exchange semester at Università di Bergamo focused on data science and machine learning.

Models on this site

MLB

Hierarchical Bayesian plate-appearance model and per-PA Monte Carlo simulator; run distributions published daily and graded live.

College football

Possession-level power ratings with calibrated margin and total uncertainty, validated walk-forward against closing lines.

NFL

Bayesian power ratings from drive-level EPA with QB and rest adjustments, backtested from 2016 with untouched holdout seasons.

NHL

Shot-quality team ratings and a Poisson goal model, with attack and defense strengths, home and road splits, and public evaluation of win probabilities against the market.