Balerion Senior Associate Aidan Daoussis sits down with Johannes Waldstein, Founder & CEO of PiLogic, to discuss AI for satellite diagnostics and sensor fusion. PiLogic is building probabilistic reasoning models for spacecraft health monitoring, onboard fault remediation, and radar-based sensor fusion in contested environments. The conversation focuses on why exact causal inference can outperform rules-based systems and machine learning in high-consequence space and defense applications.
Timestamped Overview
00:00 – Introduction to PiLogic and Johannes Waldstein’s background, including the company’s focus on probabilistic inference for space systems.03:14 – Current product areas: satellite diagnostics, onboard remediation, and sensor fusion for radar and tracking applications.05:17 – Concrete examples of anomaly detection, model building from spacecraft data, and interpreting noisy telemetry.09:25 – Macro trends driving demand in space intelligence, including sovereign space capability, connectivity, launch cost reductions, and onboard inference.12:30 – Why PiLogic’s approach differs from AGI claims and how its models are built for narrow, physics-based problem sets.14:51 – Discussion of moat: automated model generation, scalable probabilistic inference, and compression to run on low-spec hardware in space.18:35 – Broader misconceptions around AI, including the limits of LLMs and the importance of matching techniques to specific problem types.22:39 – How the system handles uncertainty, sparse data, noisy inputs, and mission-specific configurations across different satellite architectures.27:14 – Whether AI can truly reason about orbital mechanics and why space problems may require specialized algorithms rather than one general model.29:29 – What is underestimated in the space market today, including the tension between legacy aerospace development cycles and fast-moving new entrants.31:47 – Audience Q&A on launch, cross-functional operations, and the advantages of Bayesian networks over neural networks for exact reasoning.34:57 – Where PiLogic is seeing traction today across AFRL, Space Systems Command, commercial satellite operators, and radar customers.36:31 – Examples of how satellites are lost, common failure modes in orbit, and where onboard reasoning could improve resilience.40:14 – Dead satellites, orbital debris, Starlink-enabled tracking data, and the balance between sensing, inference, and connectivity in future space systems.42:50 – Adversarial satellites, space defense scenarios, and the current limits of unclassified defensive options in orbit.45:06 – Closing discussion on space insurance, risk analytics, and final takeaways on AI technique selection and PiLogic’s customer focus.
