Physical Systems Modeling · Probabilistic Reasoning · Scientific Software
Fifteen years across power systems, weather and environmental modeling, forecasting, engineering simulation and production software. I am drawn to problems where physical models, imperfect measurements, uncertain operating conditions and computational limits all have to be brought together to support an engineering decision.
A physical system, incomplete information, and computation — turned into an engineering decision.
Models that combine physical constraints with uncertainty in operating conditions, component states, parameters and measurements.
Probabilistic inference · uncertainty quantification · risk analysis · Bayesian reasoning · scenario analysis
Modeling and software for transmission systems, dynamic line ratings, grid operations, renewable energy, forecasting and reliability.
DLR/AAR · transmission · weather–grid interaction · contingency risk · EMS integration
Physics-based modeling of interconnected systems: thermal, fluid, environmental and energy.
Conservation laws · thermal systems · CFD · hydrology · physical simulation
Turning mathematical and engineering models into software that engineers and operators can actually use.
Python · Java/Spring · TypeScript/Angular · MATLAB · PostgreSQL/PostGIS · AWS · Docker
Fifteen years of physical modeling and software, of which TERRA is the most recent strand.
Probabilistic inference over physical systems: state estimation, competing-explanation diagnosis, risk and measurement design, carried across power, thermal, water and storage domains.
Independent research and software · details below
Weather modeling, conductor thermal models and the operational software around them: ratings computed from field sensors and forecasts, delivered through services and dashboards utilities can act on.
Transmission · DLR/AAR · sensor integration · cloud and on-premise delivery
Solar irradiance from satellite imagery, electric-load and solar-power forecasting, river routing and flood emergency tools, evapotranspiration datasets for groundwater models, and radiation heat transfer.
Meteorology · hydrology · statistical forecasting · scientific computing
Featured project
TERRA is a computational framework for probabilistic reasoning about physical systems. It combines physical simulation, uncertain observations and probabilistic inference to estimate system state, weigh competing explanations, quantify risk and identify the measurements worth making.
An ongoing independent research and software project.
Traditional analysis asks
What happens under this scenario?
TERRA explores
Given what we know now, what states of the system are plausible, what could fail, how likely are those outcomes, and what observation would help us understand the system better?
TERRA represents uncertain physical systems as graphical probabilistic models coupled to physical constraints and engineering solvers, so the conservation laws and the uncertainty live in one structure.
What can fail, how likely is it, and what are the consequences?
Which component, parameter or measurement best explains what we observe?
What is probably happening where we cannot measure directly?
Which uncertainties contribute most to system risk?
What should we measure next to reduce uncertainty most?
How would risk change under a different operating condition or intervention?
Probabilistic contingency analysis and physical risk assessment under uncertainty in weather, equipment availability, operating state, topology, ratings and measurements.
Probabilistic reasoning over interconnected power, cooling, thermal, compute and sensor systems.
Where the physical system meets the atmosphere and the uncertainty is shared across assets — the thread running through my transmission and forecasting work.
The same machinery has been carried into water distribution, district heating and cooling, battery storage and chiller plant, each with its own papers.
Each study is runnable, with its numbers read from committed results.
Data-center cooling diagnosis
11 % → 57 % posterior probability of the correct diagnosis
A chilled-water loop serving about 860 kW of server heat looks far less efficient than it should, and twelve explanations — real faults and lying meters alike — account for it equally well. On the first five noisy readings the true cause, a return-temperature sensor reading low, ranked only third. TERRA chose the one follow-up check worth making, an independent load heat audit; after it the true cause ranked first.
Bayesian diagnosis · sensor uncertainty · experiment design
Sensor diagnosis
| Method | Bad sensor correctly identified |
|---|---|
| Alarm thresholds | 0 / 4 |
| Peer comparison | 2 / 4 |
| TERRA | 4 / 4 |
Four scenarios, each corrupting one channel by too little to trip its alarm, against a truth whose physics deliberately differs from the model used to recover it. Beyond finding the lying channel, the reconstructed replacement value beat the bad reading every time, cutting its error by 26 % to 85 %.
Latent-state inference · virtual sensing · model misspecification
Transmission
Combinatorial uncertainty on standard power-system test networks: how graphical structure, physical separators, probability-ordered enumeration and selective physical simulation trade computational cost against the accuracy of the risk estimate.
Power flow · N−k contingencies · rare-event risk · graphical factorization
Probabilistic Modeling of Boundary-Exchanging Physical Systems — an evolving technical text documenting the mathematical framework behind TERRA, with working domain editions for power systems, data centers, water networks, district energy and battery storage. Free to read; every number in them is generated from a committed result.
Machine-learning and inference systems over weather, terrain and field-sensor telemetry, for real-time decisions on sensor systems and power-line operations.
Environmental data modeling, forecasting and decision systems for energy and infrastructure: data ingestion, the modeling itself, product direction and client delivery.
Architecture and development of simulation and modeling platforms for renewable energy and electric transmission, centered on weather-informed dynamic line ratings.
Environmental and hydrologic modeling for infrastructure and water-resource projects: river routing and flood emergency forecasting, evapotranspiration raster datasets for groundwater models and water-demand estimation, and web tools for planning water programs.
Energy forecasting algorithms and software, and the pipelines that pulled real-time and forecast weather from public sources into operational systems.
Radiation heat-transfer methods in Fortran and C++, extended to higher accuracy at lower computational cost; before that, solar irradiance forecasting from satellite imagery, with the optimization, image processing and statistical models it needed.
| Physical modeling | Power systems, thermal systems, hydrology, CFD, heat transfer, weather |
|---|---|
| Probabilistic modeling | Bayesian inference, graphical models, uncertainty quantification, Monte Carlo, probabilistic risk |
| Scientific computing | Python, MATLAB, C/C++, Fortran |
| Software engineering | Java, Spring, REST APIs, TypeScript, Angular |
| Data | PostgreSQL, PostGIS, time series, geospatial data |
| Infrastructure | AWS, Docker, Linux, Tomcat |
| Energy | Transmission, DLR/AAR, weather integration, forecasting, EMS |
Certifications — PMP (project management) · iSAQB CPSA-F (software architecture, foundation level) · CompTIA Security+ (security fundamentals) · ISO 27001 Lead Implementer (information security management systems)
I am an engineer and software developer interested in computational problems at the boundary between physical modeling, probability and decision-making.
Over fifteen years my work has moved across solar forecasting and satellite imagery, radiation heat transfer, hydrology and flood forecasting, energy forecasting, transmission and dynamic line ratings, and the software that carries all of it. The common thread has been building models and software that turn incomplete physical information into useful engineering decisions.
More recently I have been exploring probabilistic physical-system reasoning through TERRA, particularly for electric grids and data-center infrastructure.
I enjoy problems that require moving between equations, simulation, data, software architecture and the engineering question that ultimately needs to be answered.
Questions about the method, corrections to the books, or work you think this fits — all welcome.