Models

Our models

Language models and agents pre-trained from scratch by Navitra, each built for one job and the people who do it. Every model runs inside its user’s own infrastructure, and the weights belong to the client.

Financial crime

Available

NaviDetect 1.0

A language model for anti-money-laundering and sanctions work.

Built to support a financial crime analyst’s judgement, not replace it. Every output is logged so a case can be reconstructed for audit.

Type: Language model, pre-trained from scratch
Built for: AML and sanctions analysts
Runs in: Your infrastructure; weights owned by you

Learn more Request access

Cybersecurity

Available

NaviProtect 1.0

A mixture-of-experts model for security teams.

Different expert networks specialise in different parts of security work. Agents built on it ask a person before any irreversible action.

Type: Mixture-of-experts language model
Built for: Security operations teams
Runs in: Your own GPUs or private cloud

Learn more Request access

Mortgage underwriting

Available for evaluation

Mortgage Agent

The agent prepares the case. The underwriter decides.

Reads the application file, checks it against your lending policy and returns a case summary with every figure traced to its source page.

Type: AI agent on a Navitra language model
Built for: Mortgage underwriters and lenders
Runs in: The lender’s infrastructure

Learn more Request the brief

Protein science

Published

NaviKinase 1.0

A protein language model for the kinase family.

149.6M parameters, pre-trained from scratch on 2.45 million sequences. Matches ESM-2 and ProtGPT2 at identifying enzyme class without labelled examples.

Type: Protein language model, 149.6M parameters
Built for: Protein scientists and researchers
Weights: Public on Hugging Face, CC BY-NC 4.0

Learn more Read the paper (PDF)

Review articles on each model’s training, evaluation and limitations are being prepared and will be listed on the Research page.

Compare

The models side by side

Model	Domain	Type	Status	Runs in	Weights	Research
NaviDetect 1.0	AML and sanctions	Language model	Available	Your infrastructure	Owned by the client	Review article in preparation
NaviProtect 1.0	Cybersecurity	Mixture-of-experts language model	Available	Your GPUs or private cloud	Owned by the client	Review article in preparation
Mortgage Agent	Mortgage underwriting	AI agent on a Navitra language model	Available for evaluation	The lender’s infrastructure	Owned by the lender	Review article in preparation
NaviKinase 1.0	Protein science	Protein language model, 149.6M	Published	Runs locally on Apple Silicon	Public, CC BY-NC 4.0	Evaluation report (PDF)

How every model is governed

Read our AI policy summary

Validation

Tested before release

Each model is tested against agreed accuracy thresholds on a validated test set, and the results go into its model card.

Fairness

Checked for unfair outcomes

A documented method checks for unfair differences between groups, before every material release and monthly in production.

Change control

Pinned versions

Production versions are pinned. Model, prompt and data-pipeline changes need validation evidence, Director approval and advance notice to clients.

Incidents

Kill switch and rollback

Clients can switch to fully human review at once, and we roll back to the last validated version. Affected clients are told within 24 hours.

Want a model trained on your own data?

Book a free 30-minute call or email contact@navitratech.com · reply within 24 hours · UK based