Reconfiguration
Models break when feeders change.
Grid foundation model
Certifiable AI for grid decisions that have to hold up in the field.
A zero-shot monitoring and control engine that transfers across changing power networks, reasons from sparse measurements, and reports when an operator can trust the result.
Topology-invariantZero-shot transferPer-instance certificationMillisecond control
Manifold
Informed
Neural
Dual-encoder
Research
The technology behind MIND FM x (Grid Agent)
arXiv:2606.00716
Graph Transfer Learning via Shared Latent Geometry: Theory and Applications
The dual-encoder result: a teacher trained on simulator physics, a student deployed on sparse field measurements, with transfer bounded by Wasserstein proximity.
arXiv:2509.08672
Universal Graph Learning for Power System Reconfigurations: Transfer Across Topology Variations
The topology-invariant backbone: one trained model transfers to unseen reconfigurations without retraining on the new grid.
Operator cockpit
Less dashboard. More verified action.
Sparse telemetry
Distribution data is incomplete by default.
Operational trust
AI must know when not to act.
Architecture
Transfer, certify, control.
The platform is a research demonstration engine for studying how transfer, certification, and feasible control can support grid operations, planning studies, software prototypes, and market-aware grid analytics.
Grid inputs
DSS / topology
Telemetry
Loads & devices
MIND FM x (Grid Agent) engineManifold-Informed Neural Dual-encoder
Transferable grid representation with per-instance feasibility checks.
Research outputs
Power-flow solve
Certificate
Control action
Embedded API
Manifold-informed transfer
Maps grid structure and observations into a shared operating geometry.
Neural dual-encoder
Keeps a deployable pathway aligned with the physics pathway used in training.
Feasible control manifold
Selects constraint-satisfying actions in real time.
Explore
Move through the site by question, not by scroll length.
Research themes
Five application themes, one underlying grid engine.
Theme 01
Utilities & Co-ops
MIND FM x (Grid Agent) for zero-shot distribution power-flow solving and cyber-resilient control on SmartDSS systems the model has never seen, with 0 system-specific fine-tuning.
Theme 02Embedded Engines
A certifiable AI engine for grid software prototypes and research services that need faster analytics inside existing study workflows.
Theme 03Transmission & ISOs/RTOs
MIND FM x (Grid Agent) for transmission operating intelligence: fast voltage prediction, contingency screening, dispatch feasibility, market sensitivity, and operator verification across changing grid states.
Theme 04Energy Markets
Physics-aware congestion and dispatch intelligence for trading, asset optimization, and market analytics teams.
Theme 05Flexible Loads
Grid-aware coordination for data centers and other controllable loads that need operational flexibility.
Validation
Evidence that reads like infrastructure, not demo ware.
100unseen distribution topologies
12 msAC OPF decision time
97xfaster than MPC baseline
93-97%zero-shot false-data localization
Research collaboration