Applied economics and cloud-scalable modeling

Petabyte is an R&D lab that provides economic analysis and develops machine learning tools
for litigation, investment, policy, and operations optimization.

Expert-led

Our economists draw on academic research in economics, statistics, and machine learning. We provide economic analysis and design models around clients' questions, data, and systems. Our R&D focuses on developing new computational economic methods and AI/ML tools.

Economic models

Economic models give white-box answers grounded in coherent economic structure. They can use contextual data and expert judgment to improve accuracy and calibration relative to AI alone. They can also offer provable guarantees of robustness, optimality, and convergence.

Cloud-scalable

Applications like forecasting, simulations, optimization, and coherent, complex decision-making can require fine-grained, high-dimensional data. Our solutions use multi-GPU and cluster-accelerated computing, and integrate with your data lakes, on premises or in the cloud.

Products & Services

Economic Analysis and AI/ML Tools for Litigation.

Petabyte provides economic analysis for litigation and related proceedings, including antitrust and intellectual property matters. We analyze market structure and conduct, estimate damages, and perform valuations. We also assess allocation methodologies, reconstruct financial results from business records, and evaluate competing economic claims. Our litigation work draws on and informs our R&D in economic modeling, eDiscovery technologies, and large-scale data analytics.

Contact Us

Economics-based Machine Learning for Private Capital Investment and Oversight.

We build scalable decision support systems, grounded in economics-based machine learning, for investment, divestment, and oversight in venture capital, private equity, M&A, and small business lending. Our data- and theory-driven approach delivers reliable, well-calibrated, and transparent models that quantify and attribute uncertainty and value for investment committees, regulatory compliance, and litigation.

Contact Us

Bayesian simulation and forecasting at scale.

CatFish is a fully customizable, multidimensional Bayesian system for simulating and forecasting count data outcomes, including product demand. It provides location-specific, consistently aggregatable and disaggregable, full joint-distribution inference with human-interpretable factors. CatFish integrates with your cloud data infrastructure so you can run probabilistic economic models at up to petabyte scale.

Learn more

Minimize High-Dimensional Risk Functions Using Adaptive Bayesian Learning.

Newsvendor minimizes high-dimensional risk functions that encode complex, real-world determinants of profit and loss. It solves modern generalizations of its eponymous 1888 mathematical problem at scale. Newsvendor can optimally place millions of inventory items into thousands of locations or maximize returns while minimizing risk across billion-dollar investment portfolios.

Contact Us

Team and Collaborators

Dr. Ed Egan

Dr. Egan is Petabyte's founder and principal economist. He received his Ph.D. from UC Berkeley's Haas School of Business and held appointments at the NBER, Imperial College London, Rice University, and Georgetown University. He co-led scalable Bayesian machine learning R&D with Prof. John Geweke in Amazon's Core AI Group.

Bayesian R&D

Prof. Emeritus John Geweke is a Bayesian econometrician, a past president of the International Society for Bayesian Analysis, and the author of more than 100 refereed papers. Dr. Garland Durham specializes in Bayesian modeling and GPU computing. He has published in the Journal of Financial Economics and the Journal of Econometrics.

Collaborating Experts

Petabyte draws on a network of economists and computer scientists, including current and emeritus faculty at universities in the United States, Canada, and the United Kingdom. Our longstanding relationships with trusted colleagues let us bring the right experts to each engagement.

Selected Federal Proposals

We made an invited submission for Phase I SBIR funding to the National Science Foundation in early 2025. Our project tests whether Bayesian machine learning can scale to solve real-world supply chain optimization problems. It combines full joint distribution demand forecasts from a spatiotemporal economic model with high-dimensional loss functions to optimize inventory placement. This approach could begin a new paradigm of industrial economic decision-making.

We applied to the Defense Logistics Agency's Emergent IV BAA in late 2024. Our proposal develops supply chain optimization tools to meet DLA's need for scalable, interpretable systems that support combat readiness. We model uncertainty and optimize materiel placement under posture, precision, and logistical constraints. The goal is to improve responsiveness, reduce cost, and empower DLA personnel to assess risk and make mission-aligned decisions across tactical, operational, and strategic levels.

The Small Business Administration's Office of Advocacy solicited research proposals from economists in late 2024. Our proposal creates a Bayesian simulation of the U.S. small business economy to evaluate the effects of federal policies and natural disasters at the county level. The project focuses on HUBZones, Opportunity Zones, and rural trends, and supports data-driven, evidence-based policymaking through causal inference and risk assessment.

Connect with us

info@petabyteeconomics.com
Website URL

www.petabyteeconomics.com