The physical system behind AI.

Current work

Four research questions.

1. 01How do gigawatt-scale AI loads join a grid that is already constrained?Grid-integrated AI factories

   How large AI loads can connect to constrained grids while maintaining resilience, reliability and power quality.

   Relevant experience — Five years on grid connections and HVDC delivery, including construction execution on a 1,250 MW line into New York City.

   • Grid integration
   • Power quality
   • Data centers
   • Reliability
2. 02What if electricity, cooling and compute were planned as one system?Energy–compute co-optimization

   How electrical, thermal and compute infrastructure can be planned and operated as one flexible system.

   Relevant experience — Portfolio strategy across data centers, grid integration and power electronics — the three budgets that are usually optimised separately.

   • Flexibility
   • Thermal
   • Workload scheduling
   • Optimization
3. 03Can one interoperable model connect design, construction and operation?Digital twins, OpenUSD & Omniverse

   How interoperable digital twins can connect decisions across AI-factory design, construction, operation and optimization.

   Relevant experience — Product and delivery work spanning design, construction and operations, where design intent is lost between disconnected tools.

   • OpenUSD
   • NVIDIA Omniverse
   • Interoperability
   • Lifecycle
4. 04What has to be true before intelligence can act in the physical world?Industrial & Physical AI

   How intelligence moves from models into factories, infrastructure and safety-critical physical environments.

   Relevant experience — Machine-vision deployment across 29 manufacturing plants — industrial AI judged on uptime, not benchmarks.

   • Physical AI
   • Industrial AI
   • Safety-critical
   • Deployment