Decision support and principled inference for partially observed systems across earth, environmental, and health sciences — determining what actions to take, what experiments to run, and what measurements are worth collecting when interventions are costly and uncertainty is unavoidable.
Website · Publications · Projects
- Decision support under partial observability — surveillance design, forecasting pipelines, intervention evaluation, resource allocation
- Model criticism & evaluation — structured observables, Pareto-optimal configuration selection, Bayesian stacking, proper scoring rules
- Scientific AI/ML — physics-embedded surrogates, lawful learning, generative model design
- Operator-partitioned solvers — IMEX/PDE operator splitting, trait-structured dynamical systems
- Bayesian latent variable models — variational PCA, posterior predictive testing, rank selection
- Operational forecasting — infectious disease scenario modeling, ensemble calibration, intervention timing
| Package | Language | Description |
|---|---|---|
| trade-study | Python | Model evaluation and decision support: protocol-driven simulators, proper scoring rules, Pareto optimization, Bayesian stacking |
| TradeStudy.jl | Julia | Julia implementation of the trade-study framework with native scoring rules and global sensitivity analysis |
| OpSystem.jl | Julia | Declarative specification language and compiler for structured dynamical systems |
| OpEngine.jl | Julia | Operator-partitioned ODE/PDE solver with explicit, IMEX, and fully implicit methods |
| Package | Org | Description |
|---|---|---|
| VBPCApy | yoavram-lab | Variational Bayesian PCA for incomplete data with full posterior uncertainty |
| pp-eigentest | yoavram-lab | Posterior predictive eigenvalue testing for signal rank determination |
| op_engine | ACCIDDA | Operator-partitioned ODE/PDE solver core (Python) |
| op_system | ACCIDDA | Declarative specification language and compiler (Python) |
| flepimop2 | ACCIDDA | Configuration-driven forecasting orchestration engine |