Robert Donohue Bayesian modeling · Optimization · Product
rfdonohue.io

I build decision tools with probabilistic modeling + optimization.

Climate-tech builder at Enverdex, and researcher focused on Bayesian methods, efficiency tradeoffs, and interpretable models.

Now Projects Writing About
Focus
Optimization → product
Turning research workflows into tools people can use.
Methods
Bayesian + MOO
Uncertainty-aware decisions and tradeoffs.
Writing
Clear + investor-readable
From papers → blog posts with figures and results.

Building now

View all

Enverdex — Praevion Optimization Engine

startup active

Simulation-driven retrofit optimization for building decarbonization, returning tradeoff-ready recommendations.

Bayesian Optimization OpenStudio EnergyPlus Decarbonization

Enverdex

Building a decarbonization optimization engine (Praevion) that connects retrofit decisions to cost + carbon outcomes with interpretable tradeoffs.

Updates Read the project →

Selected research

View all

GP vs SVM vs Ridge — Sample Efficiency vs Compute

academic completed

A comparative study of accuracy and learning behavior under varying sample budgets and runtime constraints.

Gaussian Processes SVM Ridge Regression Sample Efficiency

Multi-Level Models — A Probabilistic Overview

academic completed

An intuitive, probabilistic framing of hierarchical models, partial pooling, and uncertainty decomposition.

Bayesian Statistics Hierarchical Models Partial Pooling Probabilistic Modeling

Discrete-Time Hazard Modeling — US Senate Toy Example

academic active

A toy example showing discrete-time hazard modeling concepts for incumbent defeat risk over an election cycle.

Hazard Models Elections Discrete Time Bayesian Modeling

Latest writing

View all
1 min read Related project →

From classification to hazard: modeling how risk evolves over time

A toy discrete-time hazard framing for election defeat risk trajectories and decision horizons.

Hazard Models Elections Probabilistic Modeling
1 min read Related project →

When is a Gaussian Process worth it? Sample efficiency vs compute

A practical comparison of GP regression, SVR, and Ridge when data is expensive and runtime matters.

Gaussian Processes SVM Ridge Regression Sample Efficiency
1 min read Related project →

Multi-level models as probabilistic systems

Partial pooling as a generative assumption: what hierarchical structure buys you and why it generalizes.

Bayesian Statistics Hierarchical Models