Analysis that can show its work
Building sandboxed agent harnesses with explicit plans, tool use, traceable execution, validation loops, and evaluation that measures more than a polished final answer.
Danny Pichardo
10+ years across healthcare data, applied AI, registry science, and baseball modeling.
At Blue Health Intelligence: LLM evaluation, text-to-SQL with retrieval over data dictionaries, guarded data-analysis agents, clinical-journey modeling, and a fine-tuned Llama 3 8B. Before that: value-based care analytics at NewWave and national registry statistics for the American Urological Association.
Outside work: the Sports Page & Scorebook, a walk-forward baseball forecasting stack, Shelves, and a 26-source reading feed.
Current build list
Agents · evaluation · baseball
Trust earned empirically
Comparing models and workflows on quality, reliability, latency, and cost, with enough observability to learn why a system succeeds or fails.
A laboratory for uncertainty
Forecasting players, modeling contact, and building tools where probabilities face a public scoreboard every day.
Selected projects
Scorebook · forecasting · agents
The MLB Scorebook →
A newspaper-inspired baseball companion with live games, box scores, standings, pitch sequences, Statcast contact-quality views, and print-ready scorecards.
Baseball forecasting stack →
A modern pipeline spanning Statcast, Retrosheet, Lahman, feature stores, walk-forward evaluation, predictive 5x5 projections, and market-oriented backtests.
Analytic-agent harnesses
Local-first systems for data exploration with sandboxed execution, structured tools, visible plans, incremental results, and rigorous evaluation.
Blue Health Intelligence ← NewWave ← AUA
Blue Health Intelligence
Lead applied GenAI and data science work across healthcare use cases, including LLM evaluation, text-to-SQL, data-analysis agents, clinical-journey modeling, and fine-tuned open models.
NewWave Telecom & Technologies
Led data science for value-based care reporting and provider-facing analytics, spanning Databricks pipelines, predictive modeling, measure calculation, and product delivery.
American Urological Association
Served as statistical and computational lead for the AQUA Registry, quality reporting, predictive research, and national-scale urologic data assets.
See the complete experience, technical toolkit, education, publications, and CV →
Selected publications & talks
Healthcare analytics and registry science.
Predicting breast cancer complications with population-based claims data →
First-authored work evaluating a clinically guided, data-driven modeling approach on MarketScan claims.
Caregiver depression and child asthma medication adherence →
Coauthored longitudinal claims research with Janet Currie examining changes in caregiver health and pediatric medication adherence.
Enabling value-based health care using modern analytics tools →
An invited talk about healthcare analytics infrastructure, provider reporting, and building useful data products.
Off the clock