Bayesian Ecological Modeling
Hierarchical generative models for phenology and species distributions.
Stan · NumPyro · JAX
Moore, M.A., Emery, N.C., and Elmendorf, S.C.
Hi, I’m Miles — Computational Ecologist
I do
- Bayesian hierarchical modeling
- mathematical biology
- high-performance scientific computing
- machine learning for ecology
- ecological forecasting
- causal inference
DOE CSGF Fellow · PhD Student, CU Boulder
I build and apply statistical and simulation-based models for studying biological systems under uncertainty. By combining Bayesian inference, machine learning, and high-performance computing, I investigate ecological processes across scales.
DOE Computational Science Graduate Fellow
Krell Institute
PhD, Ecology and Evolutionary Biology
University of Colorado Boulder
Long Term Ecological Research
LTER Network
Tools I use
Technical Expertise
Why people come to me
-
Simulation & Mechanistic Modeling
I develop mechanisticly explicit models of biological systems. When the question is what happens if the system changes, a model that only interpolates the data you already have cannot answer it.
-
Statistical Consulting for Complex Data
I leverage my expertise to help collaborators pull meaningful (causal) insight from their data. When data are especially messy (nested and crossed effects, measurement error, censoring, missingness that is not at random) I deliver thoughtful solutions to estimate quantities of interests with transparent and rigorous uncertainity around them. I work with biomedical engineering studies and ecological field campaigns turning domain expertise into trustworthy statistical models.
-
Decision Analysis Under Uncertainty
Conservation/Restoration, economic, and agronomic questions whose answer is an action rather than an estimate. I carry the whole posterior distribution through a decision-relevant loss function (e.g. planting dates, treatment thresholds, input and irrigation schedules) so that a recommendation arrives with the cost of being wrong attached.
-
Scaling to High-Performance Computing
Scaling simulations & statistical models to leverage high-performance scientific computing. Vectorised and GPU-backed inference, MPI and OpenMP for simulation, all shipped in a reproducible environment like a Docker container.
-
Agentic AI & Workflow Automation
Agents that handle the scaffolding around a modeling project rather than the modeling itself (i.e. launching fits, sweeping configurations, gathering diagnostics, drafting a preliminary report). I build them for my own work, including an agent that walks a Stan model I've written through the full Bayesian workflow. I deliberately quarantine agentic automation into janitorial and peripheral tasks so my collaborators and can take full responsibility for the core scientific software.
Research
My work sits at the intersection of ecology, statistics, and computation. I am developing methods that handle the complexity and uncertainty inherent in natural systems.
ML for Ecological Inference
Neural networks + kernel methods paired with mechanistic models.
Python · PyTorch
High-Performance Simulation
Supercomputing for large-scale biological simulations.
Julia · MPI · Fortran
Uncertainty Quantification
Probabilistic solutions for complex systems (ODEs, Transcendentals).
Selected work
Clark, E.I., Moore, M.A., Khoury, C.K., Hulke, B.S., Kane, N.C. and Elmendorf, S.C.
Soil moisture and active layer thickness across Alaska, USA and Northwestern Canada
Moore, M.A., Hoy, E.E., Clayton, L.K., Schaefer, K., Bourgeau-Chavez, L.L., et al.
Recent
This semester: graduate-level Bayesian statistics in Applied Math and a deep dive into neural networks and transformers.
Reflections on a rigorous semester covering numerical methods and high-performance computing in C++.
Clark et al. 2025 — 46 years of sunflower multi-environment trials reveal how climate drives optimal flowering phenology.
Toolkit
Scroll sideways... there's more to the right!
Probabilistic
Stan
Hamiltonian Monte Carlo for the hierarchical models most of my work rests on.
Language
Python
The everyday language — analysis, pipelines, and anything with a deadline.
Probabilistic
NumPyro
Vectorised, GPU-backed inference when a model outgrows a single chain.
Numerical computing
JAX
Autodiff and JIT compilation underneath the sampling and the surrogates.
Language
R
Ecological data wrangling, spatial work, and figures meant for print.
Machine learning
PyTorch
Neural surrogates and kernel methods paired with mechanistic models.
Language
Julia
Differential-equation models where speed and expressiveness both matter.
Parallelism
OpenMP
Shared-memory scaling for simulations too big for a single core.
Language
Fortran
Reading, extending, and coupling the legacy kernels ecology still runs on.
Reproducibility
Docker
Pinned environments so a result computed today recomputes next year.
Reproducibility
Git
Version control and CI for papers, packages, and this site.