A.I. Factory
A multi-intelligence, AI-enabled Operating System to address intelligence problems
A Proprietary framework to optimize problem solving
The problem-centric approach starts from the intelligence problem, and divides it into a list of systems that decomposes the actions to take:
- What do we need to watch?
- Where do we need to watch?
- How, using which data sources?
Answering those questions requires a strong tradecraft to come up with a combination of sensors, free from the silos that used to separate them.
A dedicated pipeline for a Multi-INT approach
The AI Factory draws and structures a variety of data coming from all intelligence disciplines (OSINT, ELINT, IMINT, HUMINT), making the tool sensor-agnostic.
Hence, one needs only to select the right combination of sensors adapted to the subsystems, regardless of the intelligence category it falls under.
The OS hence displays a modular and comprehensive solution to all situations, whether at the strategic, operative or tactical level.
Relying on state-of-the-art AI to provide reliable and comprehensive information
Unified database
The unicity and modularity of the OS relies on its unified database of structured data, where siloes are broken.
Combined with our experience in data engineering we designed a unique defense oriented data hub, including raster, vector and structured data to ensure our customers that our algorithms are trained only on the most relevant data and tested on the most realistic use cases.
Algorithm factory
Delivering solutions to the most demanding clients in the world, drive us to reach top performances on very specific applications where off-the-shelf neural networks architectures reach their limits.
Based on neural architecture search and a proprietary Deep Learning framework, not only have we industrialized the algorithm creation process to create new detectors faster than any other AI team in our sector, but we can also adapt the tool to the most advanced use cases in few clicks, fusing together several state of the art components automatically to achieve the best performances for our clients.
Overcoming scarcity
Our algorithms are specifically built to work on classified data, data we are not able to extract or even look at.
To ensure that all detectors provide the best possible performances, even on sources that are not available for training, we have developed proprietary data generation technologies based on the latest generative and 3D techniques. Coupled with other frugal approaches such as advanced distillation and smart active learning, we cope with the scarcest use cases and deploy tools for clients to improve our solution autonomously in full confidentiality.
For a powerful Man-Machine Teaming
The OS has been developed to elevate human capabilities with better valorized data, leaving analysts to focus only on high added-value tasks.
Cueing analysts
The combination of top-class change-detection and object classification artificial intelligence algorithms enable analysts to get a deeper analysis of the situation, tipping and cueing them to the events that require their expertise.
100% of the available data is now computed, redirecting analysts’ attention towards the 1% of events of interest or abnormal behaviors that have been flagged or marked by the machine.
Proprietary knowledge base
The knowledge base — which is in constant communication with the OS — encapsulates the tradecraft of Preligens’ workforce (comprised of 15% of former intelligence professionals coming from all disciplines).
It provides Preligens with the legitimacy to break down intelligence challenges into a series of comprehensive subsystems, not only to find the right combination of sensors to address them, but also to make sure that all application software are built in a comprehensive way for all situations, whether strategic, operational, or tactical.
Seamless integration
The AI Factory is deployable on premise, fusing sovereign data and openly available information in a multi-intelligence approach, and acting autonomously on this unknown environment.
For this reason, it completely relies on an evolutive and iterative approach with the analysts, whose tradecraft is key to correct and adapt the tool’s behavior.
The Operating System offers a user-intuitive human-machine-interface for human review and interaction with models, at all times, enabling users to build a tower of knowledge about the past, comprehend present events faster than their adversaries, and anticipate potential future threats like never before.