Product Workflows Customers Pricing Security Docs Sign in Book a demo

Litigo product · Prediction

Verdex

Case-outcome and settlement prediction, trained on millions of dockets. Settlement bands, motion-grant probability, and time-to-resolution — each with a confidence interval and the comparable matters behind it.

Book a demo → See pricing

Predictive AI AI analytics Scenario simulation

Primary hardware · H100 / H200 / DGX

The problem

Litigators price cases from memory and a handful of analogous matters. Partners over-litigate weak positions, under-settle strong ones, and can't hand clients the data-backed forecast they now demand.

What Verdex predicts.

Verdex trains jurisdiction- and judge-specific models on millions of dockets and your firm's own matter history — then shows its work.

Settlement bands

Damages, with intervals.

Settlement-value and damages forecasting with an explicit confidence interval — not a single number pretending to certainty.

Motion odds

Grant probability by judge.

Motion-grant probability conditioned on the specific judge and jurisdiction, not a national average.

Timing

Time to resolution.

Duration and litigation-cost projections for budgeting and client planning.

Explainability

Similar matters.

Every prediction opens a panel of the comparable matters behind it.

Simulation

Thousands of rulings.

Scenario simulation across possible outcomes to stress-test strategy.

Built on the NVIDIA stack.

Layer	Technology	Why it's here
Hardware	H100 / H200 · DGX · RTX	Multi-GPU training over millions of dockets; analyst-side batch scenario runs.
Libraries	CUDA · cuDNN · RAPIDS (cuDF/cuML)	GPU feature engineering and gradient-boosted / embedding models over large tabular docket data.
Serving	TensorRT · Triton	Low-latency, sub-second prediction endpoints.
Platform	NVIDIA AI Enterprise · NIM	Packaged, supported deployment — including on-prem for firms and insurers.

Training gradient-boosted and embedding models across millions of high-dimensional records — plus Monte Carlo scenario simulation — is intractable on CPU at firm scale. RAPIDS + CUDA cut training cycles from days to hours.

Why it holds up.

An owned data flywheel

Each resolved matter labels the next model checkpoint. The judge- and jurisdiction-conditioned model is an artifact your firm owns — not a call to a hosted API.

Judge-level granularity

Proprietary training on millions of dockets plus your private outcomes produces granularity incumbents can't match with rules or generic APIs.

For firms and risk buyers

A Litigo add-on module for litigation partners, plus an insurer / litigation-finance tier priced per seat and per forecast.

Confidence, not vibes

Every forecast carries a calibrated interval. Below threshold, Verdex abstains and surfaces the gap rather than guessing.

Part of the ecosystem

Verdex consumes the same docket corpus and tagged evidence as Discovo, Deposa, and Citel. Its predictions feed the Litigo Partner agent's triage — one corpus, one model layer, one audit trail.

Forecast your next matter.

A working demo, on real dockets, in 30 minutes.

Book a demo →