Bio
Adrian E. Radillo, Ph.D.
adrian DOT radillo AT protonmail DOT com
LinkedIn •
GitHub •
adrianradillo.org
Senior Data Scientist and Full-Stack Statistical Modeler with over 5 years of enterprise
experience deploying mathematical models into production environments and architecting
end-to-end applications. Expertise spans rigorous Bayesian statistics, predictive modeling,
and full-stack web architecture (Python, FastAPI, NextJS, Linux VPS).
Proven ability to lead cross-functional offshore engineering teams and drive complex system
modeling. Seeking a Senior Data Scientist or Backend Engineering role within enterprise
infrastructure to optimize system reliability, network data modeling, and performance.
Technical Skills
Python (Advanced), JavaScript/TypeScript, SQL, R.
Linux Server Administration, Apache Service Management, Virtual Private Servers (VPS), FastAPI, Git, CI/CD pipelines.
NextJS, React, Modern Web App Architecture.
Bayesian Decision Theory, Statistical Inference, Probabilistic Modeling, Timeseries Analysis, Temporal Modeling.
Agile Project Management, Offshore Team Leadership, Cross-Functional Technical Direction.
Professional Experience
- Architected and deploying a full-stack Knowledge Management and Debating web application serving live traffic via Apache on a Linux VPS.
- Engineered the backend using FastAPI and Python, ensuring low-latency data retrieval and robust API integration.
- Developed a highly responsive frontend using NextJS and modern JavaScript frameworks.
- Managed end-to-end infrastructure, including server provisioning, multi-domain management, network security, and database architecture.
- Conducted continuous independent research into mathematical psychology, metaphysics, and complex systems modeling. https://doi.org/10.5281/zenodo.19058514
- Developed the Python ML backend of a timeseries anomaly detection web app, now being used across 3 regions for Business analytics and financial risk management.
- Designed and developed a Root-Cause Analysis module for KPI timeseries anomalies relying on XGBoost predictive modeling and SHAPley values.
- 2+ years of heavy cross-functional communication with UI, Business stakeholders, and IT teams to deliver two high-impact analytics products into production, while securing Business and leadership buy-in. The total company budget for both projects was around USD 30M.
- Developed and maintained a Django web app for human labeling of webtext. The collected labels were used to fine-tune a Bert model used in production. The whole system is saving the company hundreds of hours of underwriting time per year.
- Setup the job and tracking infrastructure on Databricks and MLFLow to run experiments for an Operations Research project. Setup the CI pipeline for IT handoff.
- 4+ years of end-to-end predictive model building for the insurance industry (GLMs, XGBoost).
- Instructed secondary students in foundational and advanced mathematics during a 6-month, purpose-driven teaching engagement.
- Designed and ran a human Psychophysics experiment investigating temporal information filtering in perceptual decision making.
Education
Selected Publications
• Adrian E. Radillo, Alan Veliz-Cuba, Krešimir Josić, Zachary P. Kilpatrick; Evidence Accumulation and Change Rate Inference in Dynamic Environments. Neural Comput 2017; 29 (6): 1561–1610. doi: https://doi.org/10.1162/NECO_a_00957
• Adrian E. Radillo, Alan Veliz-Cuba, Krešimir Josić, and Zachary P. Kilpatrick. Performance of normative and approximate evidence accumulation on the dynamic clicks task. Neurons, Behavior, Data analysis, and Theory, 2, September 2019. doi: https://doi.org/10.48550/arXiv.1902.01535
• Radillo, A. (2009). L'expérimentation de l'utilisation des jeux vidéo en remédiation cognitive. Enfances & Psy, 44(3), 174-179. https://doi.org/10.3917/ep.044.0174
• Radillo, A., & Virole, B. (2010). Cyberpsychologie: Remédiation des apprentissages, pratiques thérapeutiques, analyse des comportements. Dunod.