AI systems
AI systems for defensive and offensive military applications
Cherev develops AI for military and national-security organizations. Our systems process intelligence, support analysis and decisions, and support missions, for defensive and offensive applications. They are built by a team with operational experience, for use under pressure by the people who carry that experience today. The Cherev Intelligence Platform is available for evaluation now; mission programs are delivered directly to customers.
Mission AI
Where Cherev applies AI
Cherev develops AI systems for military and national-security customers across defensive and offensive applications. The work is directed by our team's operational experience and by direct contact with units facing current threats, and it is engineered for the edge: degraded connectivity, exhausted users and decisions with no time to spare.
- 01
Intelligence processing, analysis and decision support
Turning large volumes of captured material, reports and imagery into ranked, structured products an analyst or commander can act on. The intelligence platform below is the productized form; customer programs adapt the same pipeline to their own data, languages and doctrine.
- 02
Mission support, defensive and offensive
AI that supports the planning and execution of military missions, defensive and offensive. These programs are delivered directly to customers; their scope, customers and status are not described publicly. Qualified organizations can request a briefing.
- 03
Simulation and training
Performance measurement and after-action review inside our training system, so that what the model observes in a drill becomes a specific correction for a specific operator.
Customer names, deployment locations, performance figures and operational details of these programs are not published. That is a security requirement, not a gap in the record: full briefings are given to qualified organizations.
Available today
Cherev Intelligence Platform
A document and imagery intelligence system. An analyst sends a PDF, a photo or a question through a secure messaging interface and receives a structured Hebrew product: translation, entities, location context, equipment identification and cross-references against the organization's own knowledge.
- 01
Translation and classification
Detects the source language of a document or image (Arabic, English, Persian, Turkish and others), translates it fully into Hebrew, preserves names, numbers and dates, and gives Arabic names in both Hebrew transliteration and original script.
- 02
Entity extraction
Pulls out people, organizations, locations, dates, equipment, financial figures and communication identifiers into a structured record, and scores the document's relevance for the organization's analysts.
- 03
Location context
For every location mentioned, the platform adds geographic context, who controls the area, relevant operational history and strategic significance, drawn from the model's regional knowledge of the Middle East.
- 04
Equipment identification from imagery
A vision model checks images for weapons and military equipment and returns type, likely model and origin, run as a second pass after the main analysis.
- 05
Metadata and telemetry
Image metadata (EXIF) is extracted and checked before processing. Files that fail the security gate are rejected without revealing the reason to the sender. Retained metadata is stored with the document for analysts.
- 06
Cross-referencing and knowledge base
Extracted entities are matched against the organization's own knowledge base and an internal reference set, so a name that appeared last month is surfaced when it appears again.
- 07
Deep research
On request, the platform synthesizes an organization's accumulated document summaries into a single research product using a larger reasoning model, and returns it to the analyst who asked.
- 08
Alerts and governance
High-relevance documents trigger an alert to the organization's administrators. Each organization has its own usage budget, audit log and administrators, managed without Cherev staff seeing the content.
- Delivery
- Secure messaging interface. Analysts submit documents, images or questions from the field or the desk; results return in Hebrew.
- Tenancy
- Each organization is isolated with its own storage, knowledge base, budget and administrators.
- Data protection
- Sensitive fields are encrypted at rest per organization. Failed documents are recorded without their content.
- Models
- Multi-model pipeline with automatic fallback. Vision, OCR and reasoning models are selected per task; model choice is a configuration, not a dependency.
- Hosting
- Cloud-hosted in the EU region today. On-premises or sovereign-cloud deployment is scoped per customer.
- Status
- Built and running. Available for evaluation by qualified organizations.
The platform runs on commercial foundation models behind Cherev's own pipeline, prompts, security gates and encryption. It is not a chatbot wrapper: the value is in the structured product, the cross-referencing and the per-organization controls. Model outputs are assistance for an analyst, not a substitute for one.
Evaluate the platform with your own material
Evaluations run inside an isolated organization workspace with your administrators. We start with a briefing and a small set of representative documents.