Hi there — I'm Alvaro Aguado. I build AI systems that make decisions at scale, and I lead the teams that run them.
I'm Global Head of Advanced Analytics, ML & Innovation at Opella (formerly Sanofi Consumer Healthcare), leading a 12-person global function with a €1.5M budget and reporting outcomes directly to the CEO and CFO. My team architected and shipped a production agentic AI platform: LLM agents on the Claude API and LangGraph extract signals from unstructured social and e-commerce data, feed Transformer and LSTM forecasters, and explain every output with SHAP. It runs across a multi-billion-dollar brand portfolio and supports B2B planning with major global retailers. I also led our Marketing Mix Modeling transformation onto Bayesian frameworks (Google Meridian, LightweightMMM), delivering +5% year-over-year media ROI.
In parallel, I work part-time as a domain expert for leading frontier AI laboratories — producing expert preference data and rigorous critiques that train models to reason correctly in data science, statistics and machine learning, and advising on how frontier models should produce industry-standard analytical outputs. I also teach graduate courses in text mining, optimization and advanced analytics at the American Public University System.
Over the past decade I've worked across pharmaceuticals, fintech and retail — Opella, Sanofi, Pfizer, Verisk/Argus, Experian and Deloitte — building everything from large-scale data-intensive systems (30B+ transactions per month) to real-time analytics on sparse and unstructured data. Recurring themes in my work: demand and innovation forecasting, marketing mix and multi-touch attribution, price elasticity, online review and social listening analysis, conjoint models for innovation, segmentation and customer lifetime value, and computer vision for manufacturing optimization.
Alongside my industry work I completed a PhD in Business Data Science at NJIT (GPA 4.0, unanimous committee approval), where my research examined how text and images jointly influence online review helpfulness and virality. The work extends the Elaboration Likelihood Model into a quantitative multimodal framework (TI-ELM), explaining how consumers form judgments through informativeness, persuasion and credibility mechanisms. It is published in the Journal of Business Research.
What I care about: getting LLM-based systems from demo to production, measuring whether they actually work (evals, conformal prediction, causal methods), and building teams that can do both.
When I'm not coding I'm taking care of my daughter Livia, or playing basketball, tennis, or any sport that may seem fun. I also have a passion for dogs and a growing hobby for mixology that's still a work in progress with a lot of errors. Thank you for visiting — I hope you find something here worth your time.
Stack: Python · PyTorch · Claude API · LangGraph · Transformers · PyMC / Stan · Google Meridian · LightweightMMM · R / Shiny · SQL · Snowflake · GCP · GitHub Actions · Power BI
