Know when two datasets truly agree
Deploy Growt's structural alignment engine under your own brand. Validate AI model transfer, compare sensor pairs, certify cross-site data equivalence — any domain where two data sources need to agree. Standard metrics miss structural divergence. Growt catches it.
Want the technical details? See how Growt works →
Where operators deploy Growt
The core question is always the same: are these two windows onto the same thing actually seeing the same thing?
AI model transfer
Does your trained model transfer to production data? Validate before you ship — not after it fails.
Sensor pair comparison
Two radars, two lidars, two cameras observing the same scene — do they structurally agree, or has one drifted?
Multi-site clinical studies
Patient data from Site A vs Site B — certify structural equivalence before pooling cohorts or training shared models.
Cross-modality alignment
Radar vs optical, Sentinel-1 vs Sentinel-2 — does the structural content of two modalities agree over the same geography?
Equipment calibration
Two test benches, two lab instruments — are they producing structurally equivalent results or has calibration drifted?
Temporal drift detection
Has your data distribution shifted between last quarter and today? Catch regime change before your model notices.
Sim-to-real transfer
Does your simulated training data structurally match real-world sensor readings? Validate the reality gap before deploying a robot trained in simulation.
Sim-to-sim comparison
Two simulators, same environment — do they produce structurally equivalent training distributions? Validate before migrating from one physics engine to another.
Domain randomisation
Does your randomised sim actually span the real-world variation your robot will encounter? Structural coverage analysis — not guesswork.
LoRA adapter & task testing
Does a fine-tuned adapter trained for one task or domain structurally transfer to new data? Validate adapter compatibility before deploying to a new base model or dataset.
Pharma & drug discovery
Cross-lab assay reproducibility, compound library transfer, multi-instrument calibration. Certify that two labs' molecular data are structurally equivalent before pooling or modelling.
Predict from any single sensor
Run all your instruments together once during calibration. Then drop any one in the field — Growt infers what the missing instrument would have recorded. Radar only? Predict what the camera would have seen.
The entire competitive landscape answers one question
“Are they different?”
MMD gives you a distance. KS gives you a p-value. Classifier tests give you an accuracy score. Every drift detector, every monitoring platform, every two-sample test — they are all detectors. Binary signal with magnitude.
Growt answers a different question:
“How are they different, and where?”
What every other tool gives you
- ✗A distance score or p-value
- ✗A drift alert when the number crosses a threshold
- ✗A chart showing the metric over time
- ✗No indication of where or why
What Growt gives you
- ✓Which regions of the reference space the new data never reaches
- ✓Which specific dimensions are driving the divergence
- ✓Per-sample anomaly scores — which inputs are genuinely novel
- ✓Prioritised recommendations with evidence, not just a number
How Growt compares
vs MMD · KS · Wasserstein
Global statistics that miss structural differences not reflected in moments. Same mean, same covariance, completely different internal structure — they won't catch it. Growt will.
vs Classifier two-sample tests
Can the classifier distinguish dataset A from B? Yes or no, with an accuracy score. It won't tell you where or why. Growt identifies which part of the structure is mismatched.
vs Evidently · Arize · Fiddler
Built for post-deployment time-series monitoring. Not designed for pre-deployment validation, cross-instrument comparison, or multi-site equivalence certification.
vs all of them
Every competitor is either an open-source library or a monolithic SaaS. None offer a franchise model. Growt is the only structural data intelligence platform built for domain expert operators.
Every existing tool tells you two datasets differ.
Growt tells you how — which regions, which dimensions, how much coverage is at risk.
Pre-deployment, not post-incident.
Compliance & regulated industries
The insight is valuable. The record is required.
In medical, pharma, finance, and any EU AI Act context, knowing your model's transfer quality is not enough. You need to prove you checked — with a timestamped, signed record that survives an audit.
Immutable audit trail
Every audit run is timestamped and stored. “On 14 March 2025, model v2.3 was audited against production sample. Oracle: 71%. Safe to deploy.” That record exists. Your compliance officer can pull it.
Regulatory coverage
EU AI Act Article 9 requires pre-deployment validation evidence. FDA expects documented distribution checks for medical AI. Financial regulators expect model risk documentation. Growt generates structured reports your customers can attach to submissions.
Drift monitoring over time
Re-audit monthly. When oracle drops from 71% to 48%, you see it before your customers do. Set alert thresholds. Fire webhooks to Slack or PagerDuty. Turn post-incident forensics into pre-incident SLAs.
The emnist case: Val accuracy held at 72%. Oracle collapsed to 0.7%. The model looked fine by every standard metric and would have shipped — the audit record caught it. That is exactly what a compliance officer wants to see: that you have a mechanism that catches what accuracy cannot.
Audit Confirm — recipient verification
Coming soonYour customer receives a model deployment. They query its record ID. They get back a cryptographically signed receipt — oracle score, verdict, timestamp, and a public verification URL — without needing access to your systems or data.
For your customers
Include a record ID and verify link in every delivery report. Your customer confirms it independently — no trust required.
For regulators
Query the audit registry by record ID, no data access needed. Like checking an FDA clearance number or a TLS certificate.
For automated pipelines
Call audit_confirm(record_id) in CI/CD. No valid audit record → fail the build before production.
Verification is free and public — like Let’s Encrypt for audit records. Operators include the record ID in customer-facing reports. Insurance underwriters and compliance teams can mandate it as a delivery condition.
Vision · Language · Action models
VLMs and VLAs: the hidden transfer risk
Each component passes its own benchmark. The vision encoder looks fine. The language model looks fine. But when they run together in production, the joint space is where transfer fails — and standard accuracy metrics never see it coming.
Vision space
Does the visual encoder generalise from training scenes to deployment scenes? Coverage gaps in the visual distribution flag camera changes, lighting shifts, or unseen object classes before they cause failures in the field.
Text / language space
Does the language module handle the vocabulary, phrasing, and domain shift your deployment introduces? A language oracle below 50% means the model is extrapolating — not reasoning — on production inputs.
Joint (fused) space
The hardest check: do vision and language features align the way the model expects when they are fused together? Two clean components can produce a broken joint space. This is the component that silently kills VLM deployments.
Growt audits all three spaces in a single call. A model is only marked safe to deploy when vision, language, and joint all clear the oracle threshold. One failing component blocks the deployment — regardless of how high your benchmark accuracy is.
What operators get
Your brand, your domain
Custom logo, colors, and domain. Your customers never see Growt.
Branded MCP tools
MCP tool names and descriptions use your vertical's language. Zero IP exposure.
Customer management
Onboard and manage your customers. Track usage, billing, and API keys.
Operator tiers
API
Usage-based
Stateless endpoints only. Pay per call, $0 idle.
- ✓Stateless Lambda endpoints
- ✓Branded MCP proxy
- ✓Account + usage API
- ✓Basic support
Team
Monthly
Session-based alignment + branded portal for your customers.
- ✓Everything in API
- ✓Session-based pipelines
- ✓Branded portal deployment
- ✓Usage dashboard
Compliance
Monthly
Full audit trail, GDPR exports, AI governance registry.
- ✓Everything in Team
- ✓Operator audit log
- ✓GDPR data export
- ✓EU AI Act registry
- ✓PDF report generation
Enterprise
Custom
Dedicated infrastructure, custom MCP profile, SLA.
- ✓Everything in Compliance
- ✓Silo DB isolation
- ✓Custom MCP tool names
- ✓VPN access
- ✓Dedicated support + SLA
Run your franchise with any AI agent
Your Operator MCP connects directly to Claude, GPT-4, Gemini — or any MCP-compatible agent. Ask it anything about your business. No dashboards to check.
What you can ask your AI agent
- ›"Which of my customers ran a RED FLAG audit this week?"
- ›"Who hasn't used their API key in 14 days?"
- ›"Show me MRR trend and top customers by call volume."
- ›"Alert me if any customer exceeds their tier limit today."
- ›"Which use cases are my customers actually running Growt for?"
What it tracks for you
- Usage analytics
Calls per customer, per endpoint, per day — with trend lines.
- Audit health
How many SAFE vs RED FLAG vs BAD MODEL results across your customers.
- Engagement gaps
Customers who haven't called in 14+ days — with suggested re-engagement.
- Use case detection
What domains your customers are actually solving. Flags out-of-scope usage.
- Tier alerts
Customers approaching limits, ready to upgrade, or at risk of churning.
Give your investors a live audit — not a PDF
The Investor MCP lets your investors connect their AI agent and run a full audit of your franchise anytime. Board-level governance, no spreadsheets.
Curious
Pre-investment evaluation
- ✓Platform growth snapshot
- ✓Public benchmark results
- ✓Use case overview
Invested
Active investor
- ✓Full MRR & usage trend
- ✓Churn signals
- ✓Top customers (anonymised)
- ✓Structural health dashboard
Significantly invested
Major stakeholder
- ✓Named customer detail (with consent)
- ✓Detailed endpoint breakdown
- ✓Board pack download
- ✓Monthly briefing agent
Board member
Board governance
- ✓Everything in Significantly invested
- ✓Vote on board resolutions via MCP
- ✓Full audit trail access
- ✓Executive alerts
Investor keys are issued via Keygen — scoped, rate-limited, and fully audited. All investor activity is logged. You see who checked in and when.
Read the full Investor MCP documentation →You're not alone — you're part of a network
Growt operators are domain experts. We bring you together — in dedicated 1:1 sessions and at our annual operator summit.
Dedicated 1:1 operator meetings
Regular strategy calls with the Growt team — your roadmap priorities, customer challenges, new vertical opportunities. Direct line to the people who build the platform.
Annual operator summit
Once a year, operators from every domain gather in person — share what's working, shape the product roadmap, and meet the people solving the same problem in different industries.
How the application works
We onboard selectively. The process is designed to make sure there's a real fit — for you, for your region, and for Growt.
Apply and describe your use case
Fill in the application form — your company, background, the domain you want to operate in, and the regions you want to cover. Tell us what problem you want to solve and for which customers.
Talk to the Growt AI agent
After submitting, you'll be connected to the Growt operator AI agent. Describe what you want to build, which domain you're targeting, and your preferred geography. The agent records your intentions and checks in real time whether your region and domain are available.
Availability check
The agent verifies that your target region and use case domain are open — no existing operator covers that territory and domain combination. If there's a conflict, it will suggest adjacent opportunities.
Invitation or direct start
If everything lines up, you'll receive an email. You can choose to book an optional intro call with the Growt team — or skip straight to activating your operator account. No meeting required.
Operator agreement
Once activated, you sign the operator agreement. This binds you to your agreed region and domain — protecting your territory from other operators, and ensuring customers in your area are served by you.
Template contracts
Draft templates — review with your legal counsel before use. Print or save as PDF from the page.
See how it works — in under 2 minutes
Five short walkthroughs — the operator journey, how auditing works, and how recipients verify what they receive.
1 min — from application form to signed territory
Becoming a Growt Operator
5 steps from application to your first customer
Try the Growt Operator Agent
Describe your use case, region, or the problem you want to solve for your customers. The agent will run a demo audit, check territory availability, and save ideas worth exploring.
Operator Referral Programme
Recruit a new operator — earn based on their results. When you bring a new operator into the Growt network, you earn a performance bonus tied to their activity. The better they do, the better the bonus for you.
Organisation performance bonuses. There are also bonuses based on how your operator organisation as a whole performs — total revenue, growth, and customer outcomes. The network rewards collective success.
Were you referred by someone?
If an existing operator or contact brought you here, enter their details below so they receive credit for the referral.
Public benchmark results
Real API results on standard ML benchmarks. Transfer Oracle detects distribution shift that validation accuracy misses.
MedMNIST
PathMNIST → 3 medical domains. BloodMNIST: 96% val accuracy but 18% oracle — 50% coverage.
View full results →CIFAR-10-C
Clean → 5 corruptions × 5 severities. Gaussian noise sev 5: 13% oracle, RED_FLAG. JPEG stays SAFE at all severities. Corruption-specific, not just severity-based.
View full results →
Ready to apply?
Start with a 14-day trial franchise — full Operator MCP access, no billing until you convert. Applications are reviewed personally. We onboard selectively.