Predict & Act, in one stack.
SUFI PREDICT™ <> NEXUS

NEXUS's prediction engine. SUFI PREDICT™'s data infrastructure. No discovery phase. Deterministic, embedded, and ready to ship.

Start a conversation See the proof
POC: Luxury Retail
POC: Hedge Fund
Backed: $255M Series A
CLASSICAL ML · 4H TUNING 89% AUC PRE-TRAINED ON BILLIONS OF TABLES
Fundamental
The Prediction Engine

Weeks of fine-tuning.
One API call. Hours to prediction.

Nexus arrives pre-trained on billions of enterprise tables. Every new row makes it sharper.

$255M raised · $1.2B valuation · Backed by Oak HC/FT, Valor Equity, Salesforce Ventures

Schema changes don't break predictions.
Missing values need no imputation.
Signals connect across enterprise tables.
Four hours of tuning, beaten in one pass.
Tabular foundation models vs. traditional ML Hollmann et al., Nature, Jan 2025 · 57 datasets
XGBoost 4h tuning 67%
CatBoost 4h tuning 71%
Tabular foundation model <3 sec 89%
Built on the same architectural class that outperformed XGBoost and CatBoost across 57 independent datasets — Hollmann et al., Nature, Jan 2025. Under 3 seconds vs. 4 hours of traditional tuning.
The Loop

Prediction meets production.

Sufi Predict™ — powered by Fundamental's Nexus foundation model — doesn't stop at a score. The Sufi Data Factory™ feeds it, and the Sufi Agentic Harness™ acts on it — inside your systems. The loop closes, and it compounds every cycle.

THE ENTERPRISE DELIVERY LAYER OPERATING SYSTEM = YOUR COMPANY OUTCOMES BECOME NEW SIGNALS · THE LOOP COMPOUNDS · LOGGED · AUDITABLE · REVERSIBLE
1 · Connect your data
Every source — legacy ERP to live web signals — normalised into clean tables Sufi Predict™ can read from day one.
2 · Generate a prediction
Pre-trained on billions of enterprise tables. No retraining. Accurate from the first row.
3 · Act on it
Agents fire inside existing systems automatically — before a human has to intervene.
Where It Runs

Playbook that scales across industries

Each vertical started with a specific operational failure. The stack addresses it through better predictions on data enterprises already collect.

● Live POC
Finance
Three decision problems. One prediction layer. No model training required.
Built for: Chief Risk Officers · Quantitative Portfolio Managers · Heads of Credit
99% AUC
Zero tuning · XGBoost peaks at 71%
Open the playbook ↓
Hedge Fund
Signal Generation

2-day signal lag collapses to T+0 — NEXUS on fund tables, analyst loop removed.

<3 sec
Prediction latency. Traditional pipelines need 12+ hrs of feature engineering.
99% AUC
On financial classification benchmarks. Zero tuning. XGBoost peaks at 71%.
Insurance & Underwriting
Claims Fraud Detection

Stale-rule fraud review shrinks from days to hours — NEXUS on claims tables, exceptions only.

20–40%
Improvement in fraud identification vs. static rules.
$7 : $1
Return for every dollar invested in fraud prevention technology.
Private Credit
Credit Underwriting at Scale

More deals, faster — NEXUS on revenue and covenant tables, no training data needed.

#1
Tabular FMs outperform all ML approaches across every credit risk dataset tested.
Thin data
Largest gains in SME loans and low-default books — where traditional ML fails most.
● Live POC
Retail
Wrong inventory, wrong location, wrong time. The data to fix it already exists.
Built for: VP Supply Chain · Head of Merchandising · CTO
91%
Forecast accuracy · vs 75% tuned XGBoost
Open the playbook ↓
Luxury & Specialty Retail
Demand & Inventory Forecasting

Overstock and stockouts predicted and avoided — NEXUS on ERP and POS tables, no ML pipeline needed.

91%
Forecast accuracy with predictive AI. vs. 75% tuned XGBoost, 63% rules-based.
60–70%
Industry ceiling today — when SKUs are modelled independently.
Fashion & Apparel
Markdown & Sell-Through Prediction

Markdown timing predicted before the window closes — NEXUS on SKU × store × week tables.

1 pass
Matches 8-hour tuned AutoML ensembles. No training run, no pipeline.
$301B
Global inventory distortion from supply chain disruption — IHL Group, 2025.
E-commerce & D2C
Returns Fraud & Demand Forecasting

Returns fraud and demand spikes — one prediction layer, NEXUS on transaction tables.

17–20%
Average e-commerce return rate — a growing share fraudulent or policy-abused.
Same tables
Fraud and demand forecasting run on the same data. Only the target column changes.
Next
Energy & Manufacturing
The sensor data exists. The maintenance model doesn't — in energy or on the factory floor. Failure is scheduled by calendar, not condition.
Built for: VP Asset Management · Head of Operations · Plant Engineering Lead
30–50%
Less unplanned downtime
Open the playbook ↓
Utilities & Grid
Predictive Asset Maintenance

Failures predicted, work orders dispatched automatically — NEXUS on sensor and telemetry tables.

30–50%
Reduction in unplanned downtime from sensor-based predictive maintenance.
4–5×
Emergency repairs cost 4–5× more than the same work done on a planned schedule.
Renewable Energy
Yield Forecasting & Turbine Health

Yield forecasts sharpened, faults caught earlier — NEXUS on generation and maintenance tables.

98.7% AUC
On energy benchmarks, zero tuning. CatBoost peaks at 85.6%.
18–25%
Reduction in total maintenance spend from predictive programmes.
Oil & Gas
Equipment Failure Prevention

Failure predicted days ahead — NEXUS on PI/OSIsoft historian tables as-is, no data science team.

4×
Unplanned downtime costs have quadrupled over five years. Reactive strategies compounding.
45–55%
Reduction in unplanned downtime events. Total maintenance costs cut by up to 41%.
Next
Healthcare
Rich tabular data across every layer — devices, claims, supply chain. Almost none of it has a prediction model on top.
Built for: Chief Medical Officers · VP Operations · Head of Revenue Cycle
40–50%
Equipment downtime reduction
Open the playbook ↓
Hospital Systems
Medical Equipment Maintenance

Equipment failures predicted before downtime hits — NEXUS on device telemetry, no patient data involved.

40–50%
Reduction in equipment downtime. 25–35% cut in maintenance costs within 12 months.
3–5×
Emergency repairs cost 3–5× more than planned maintenance on MRI, CT, surgical equipment.
Pharma & Med Device
Manufacturing Line Reliability

Batch failures predicted before they occur — NEXUS on sensor tables, no data science team.

Least data
Tabular FMs improve most where data is scarce. No historical failure examples needed to start.
4–8×
Unplanned production stops cost 4–8× more than scheduled — amplified by batch value and regulatory risk.
Health Insurance
Claims Fraud & Risk Scoring

Fraud flagged before payout, high-risk members surfaced early — NEXUS on claims tables.

20–40%
Improvement in fraud identification vs. static rule-based review systems.
$7 : $1
Return per dollar invested. One model covers fraud detection and member risk scoring.
Ecosystem Partners

Acceleration inside Google, Oracle & AWS.

Deploy Nexus in your native environment — Google Cloud, Oracle, or AWS SageMaker JumpStart.

Google Cloud
Google Cloud
Premier Partner
+
Oracle
Partner Network
+
AWS
SageMaker JumpStart · Jun 2026
Deploy
Customer Activation
NEXUS deployed in production — Data Factory, Predictions & Agentic Harness, inside Google and Oracle accounts.
Distribute
Marketplace Listing
NEXUS listed on GCP Marketplace and OCI Marketplace. One-click deploy.
Integrate
Native Model Embedding
NEXUS callable inside Vertex AI Tabular Workflows and Oracle Machine Learning.
What This Becomes

From data to decisions to an operating system.

Three phases. Each builds on the last. By Phase 3, the organisation isn't using AI to answer questions — it's running functions through it.

1
Phase 1
Insights
The Data Factory brings data — structured & unstructured. Nexus reads it and generates predictions.
What it feels like: For the first time, data collected for years produces decisions rather than reports.
2
Phase 2
Actions
The Agentic Harness picks up the Insights layer and acts on it. Every action feeds back into the data layer — outcomes become new signals. Predictions sharpen every cycle.
What it feels like: The organisation stops reacting to events that already happened. It starts acting on events that are about to happen.
3
Phase 3
Operating System
Multiple action loops — across finance, supply chain, operations, workforce — compose into something larger. Functions converge into an Operating System for your business.
What it feels like: The organisation does not just use AI. It operates through it.
Start here

Ready to close the loop
on your enterprise data?

We start with your data. A live scored pipeline in six weeks, or a clean exit. No obligation beyond the first proof point.

Start a conversation