JF Advanced Analytics delivers custom, built-to-order tools and analyses for complex prediction and decision-making. We provide on-demand senior data science that cuts risk, informs and accelerates decisions, and increases returns. We know the real world runs on probabilities—not absolutes—so we turn uncertain, messy, and limited data into clear, confident decisions.

• Uncertainty-aware modeling & inference: Bayesian and frequentist UQ; epistemic/aleatoric decomposition; calibrated predictive distributions; decision thresholds; active learning.
• Operations research & decision science: Stochastic scenario analysis; risk–cost–benefit analysis/multi-criteria decision analysis; robust/stochastic optimization; sensitivity analysis; Monte Carlo simulation.
• Time-series & sequential ML: Anomaly/outlier detection; pattern recognition; gradient boosting and traditional regression.
• Physics-based & engineering modeling: Finite-element analysis; design of experiments; surrogate and reduced-order modeling; model verification, calibration, and validation.
• Production-grade delivery: CI/CD pipelines; reproducible, versioned releases; containerized services (Docker); workflow orchestration (Prefect); observability and testing.

Owner and principal consultant Josh Fody is a senior data scientist, former NASA research engineer, and U.S. Navy veteran with 12+ years building advanced computational models and data-driven solutions. He holds a Master of Science in Mechanical Engineering and completed a year long Artificial Intelligence Academy. His award-winning work has been patented, licensed, and featured in dozens of technical publications and industry presentations.

Risk & model audit:
Diagnosed failure modes in an underperforming model; delivered comprehensive evaluation including fixes.
CI/CD data pipeline:
Built and deployed a scalable, containerized data pipeline (Prefect • Streamlit • SQLite • Docker) with versioned releases.
Synthetic data generation:
Estimated a posterior over custom mixture model parameters for (near) Markov increments via Bayesian MCMC; drew top-n posterior samples with likelihood weighting and propagated increments from the latest value to generate realistic forward paths and horizon-wise predictive distributions.
Manufacturing process UQ:
Optimized process parameters and quantified sensitivity for ultrasonic welding of thermoplastic composites using Gaussian Process Regression on noisy, sparse data.
Reliability under scarce data:
Developed a stochastic/UQ model to predict long time-horizon component life, for an aerospace application from limited test data, including calibrated uncertainty bounds.

Josh Fody
Owner and Principal Consultant

Data Science | Machine Learning | Uncertainty Quantification

Based in Florida, USA | Serving clients globally

+1 (414) 426-8109    josh@jfadvancedanalytics.com