California-based startup PiLogic Inc., which specializes in high precision, expert-guided artificial intelligence for the defense and space industries, announced Thursday it has raised $4 million in seed funding.

Scout Ventures and Seraphim Space led the round, with participation from Sovereign’s Capital, Flex Capital, FN Fund and angel investor Gokul Rajaram.

PiLogic’s platform design is building “exact AI,” a model architecture that is mathematically precise, faster and more accurate than generative AI large language models and uses dramatically less computing power. Due to this developmental consideration, the company’s models can be run without requiring powerful graphics processing units and on-board satellite systems.

The company explained it uses a unique approach, which combines multiple AI techniques, including probabilistic inference, search and machine learning. Its models incorporate expert-driven, first principal designs directly into their decision-making processes.

Founder and Chief Executive Johannes Waldstein said the potential of exact AI architecture cannot be understated for industries where precision, real-time inference and small footprint are important. “We believe that many of these critical applications, especially in defense and space, can’t feasibly be addressed by gen AI,” he said. “And that is what we are here to solve.”

PiLogic has released two models to address these needs for space applications.

The company’s first AI system, an Electrical Power Systems Diagnostics Model, allows satellite operators to analyze and predict the health state of components using onboard sensors, in addition to providing real-time action to correct problems. It can be implemented both onboard satellites, due to its small footprint, or run on the ground using radio telemetry. PiLogic said it has validated the model’s effectiveness through work with the National Aeronautics anSpace Administration.

A Radar Model provides tracking and prediction of objects in 3D space. The company said it can distinguish between different types of objects such as airplanes and drones, and predict when objects will cross paths or make close approaches. The model can be adapted to use outputs from sonar, global positioning satellite systems and lidar.

The startup said both models provide significantly better benchmarks than competing rules-based techniques for diagnostics and detection in similar environments.

“PiLogic’s approach to high-stakes decision-making on the edge has the potential to transform the space industry,” said Andre Ronsoehr, partner at Seraphim Space, a venture capital firm that specializes in space technology startups. “It is a sector still rife with rules-based systems and humans in the loop, often for the lack of viable alternatives.”

Future uses for the exact AI model architecture include supporting autonomous flight, automatic threat prediction and response, diagnosing and predicting electrical power system failures for aircraft and spacecraft, radar tracking and fusing diverse sensor readings.

Waldstein said the seed funding round will allow the company to scale its team and expand the core product to support early customers facing mission-critical challenges within the aerospace and defense industries.

Image: Pixabay