The trust layer for grid AI

Real grid AI
doesn't guess.

Modern grids are too complex to trust by hand. Tessera is the foundation beneath every grid AI — finding the errors others miss, validating every answer, and refusing to speak when the proof is missing.

Built on IEC 61968 / 61970 CIM
The problem

Every grid AI runs on a model nobody has systematically validated.

20%
of distribution engineering time is spent troubleshooting grid model errors — one day a week, per engineer, gone (EPRI).
$200–400M
spent each year by utilities on manual CIM configuration. No existing platform has displaced it. The work is done by hand.
Zero
grid AIs deployed today run on a continuously validated distribution model. Every one is silently exposed to the model underneath.
Where we sit

The data plane beneath the grid stack.

Every function in the utility stack — from real-time control to AI decision support — depends on a validated network model. The data those functions need is scattered across many source types and vendor systems. Tessera sits between them, rebuilding a validated CIM graph that every function above can act on with confidence.

Above. Every grid function gets measurably better — not replaced. The systems you already run keep running.
The middle. Validation, refusal, and provenance happen here — once, for everything above.
Below. Your source data stays where it lives. Adapters, not migrations.
What we do

Three pillars. One trusted source of grid truth.

Tessera is the layer between your raw GIS, asset, and SCADA data and every system that depends on it. It rebuilds the network model continuously, validates it across five independent layers, and grounds AI in data that has been proven correct — not just received.

01 — Ingestion

AI-powered ingestion

Any source utilities actually run — GIS, SCADA, CAD, PDF schematics, satellite imagery, billing, public datasets — mapped into a CIM-conformant graph with vendor-aware schema understanding. Synthetic feeders for utilities without complete GIS. Months of manual ETL collapse to days.

02 — Validation

Continuous trust gate

Five-layer validation runs on every write — schema, topology, physics, twin-sandbox, provenance. Tiered scheduling from sub-second to nightly. Defects surface with severity, suggested fix, and a clean audit trail. Engineers approve, the model heals.

03 — Grounded AI

AI proposes. Logic decides — or refuses.

When the validated graph can substantiate an answer, Tessera grounds it and cites the records. When it cannot, Tessera refuses — explicitly, not silently. No quiet hallucination, no confident guess. Verdicts are byte-reproducible, replayable, and auditable.

Trust architecture

Five layers between raw data and decisions.

Every fact in Tessera is signed at five layers of validation before it can be acted on. Nothing leaves the platform without provenance you can audit. And when even the validated graph cannot prove an answer, the platform refuses to give one — by design, not by override.

L1 CIM grounding Every claim resolves to a stable network-model identity (mRID).
L2 Topology Schema and connectivity rules per IEC CGMES 3.0 — types, references, phasing.
L3 Physics Power-flow and electrical-law checks — voltage, balance, KCL/KVL.
L4 Sandbox Proposed corrections run in a digital twin before they reach the live model.
L5 Provenance Every answer cites the records, scores, and sources it was grounded on.
The platform

An infrastructure layer — not another application.

ADMS, DERMS, digital twins, AI copilots, interconnection studies — every system running on your network model is only as good as the model itself. Tessera sits beneath them, continuously rebuilding a validated CIM graph from any data you have, on any deployment posture you need.

Visual workspace, conversational agent — same graph underneath.

Engineers who think in pipelines drag and connect nodes. Engineers who prefer conversation talk to an agent grounded in the validated graph. Both drive the same CIM data plane. Both are constrained by the same trust gate. Neither can act on data the platform hasn't proven.

GIS export
CIM map
Trust gate
Publish

Additive to your stack — not competitive with it.

Every ADMS, DERMS, GIS, and grid AI tool depending on a validated network model gets measurably better results with Tessera underneath. The validated graph is the contract. Everything above and below is your choice — and your existing investment.

Stripe sits beneath every payment app.
Twilio sits beneath every messaging app.
Snowflake sits beneath every analytics app.
Tessera sits beneath every grid AI app.
Built for the control room

Engineering choices that matter to people who run the lights.

Tessera is built to the same constraints utility procurement teams already enforce — air-gapped where it has to be, deterministic where it counts, and provably grounded everywhere it touches AI.

Air-gapped capable

Runs in an isolated network on as little as 4 GB of RAM. NERC CIP-005, CIP-010 and CIP-013 compliance is maintained, not waived. Cloud, on-prem, hybrid and air-gapped — same product, same code path.

Deterministic verdicts

Every approval is byte-reproducible. Replay a verdict from six months ago and get the same answer, with the same audit trail. No model drift in the decision layer — by construction.

Sub-millisecond traversals

Graph-native persistence with native delta versioning. Operators see the full network state in the time it takes to refresh a screen — no ETL window, no overnight rebuild.

Network effect

Every utility's validated corrections compound accuracy for peers running the same vendor stack. The longer the network operates, the harder it is to catch up — and the more your investment compounds.

Why "Tessera"

A trusted token. A tessellating tile.

Tessera hospitalis

The Roman trust token.

In ancient Rome, a tessera hospitalis was a small clay tile broken between two parties — proof of trusted identity across strangers' thresholds. Whoever held the matching half was, by definition, to be trusted. That's the bond Tessera offers between your raw data and the systems that depend on it.

Tessellation

The tile that scales.

A tessellation is a pattern of discrete tiles that interlock perfectly at every scale — fractal, modular, gap-free. That's the architecture: small, validated building blocks composed into a national grid model without losing integrity at any boundary.

Trusted token. Tessellating tile. Scaling pattern.
Loose pieces become a mosaic. A mosaic becomes the grid.
The team

Built by hands that have run the lights.

Tessera is shaped by engineers and operators with a combined century of utility OT experience — across distribution control rooms, ADMS and DERMS deployments, CIM standards work, and grid software at the world's largest utility software vendors. We have spent careers fixing the data layer the hard way. Tessera is the platform we wish we'd had.

Detailed team bios available to qualified utility partners and investors on request — get in touch.

Request a demo

See your network model — validated.

Tell us about your utility and we'll come back within two business days with a 30-minute working session.

We can run Tessera against an extract from your live GIS or ADMS in advance of the call — so the conversation is grounded in your actual data, not a generic walkthrough.

  • Live three-layer DQ run on your network
  • Findings report with severity and suggested fixes
  • No-cost pilot path for design partners