Peter Schneider

Leading AI/ML projects focused on deep learning in areas that include computer vision, synthetic aperture radar (SAR), Large Language Models (LLMs), graph neural networks (GNNs), temporal data, RF and signal processing.

• Principal Investigator for Strike Aligned AI Research, including work with LLMs, multimodal LVLMs, and agents.
• Complex-valued neural networks better exploiting both magnitude and phase in SAR and optical data; successful SAR ATR demo deployed on Triton.
• GNNs with heterogenous graphs for various applications, including for multi-view multi-object tracking.
• Hyperbolic neural networks improving performance with hierarchical data.

Deep learning with SAR imagery.

• Research and development of complex-valued neural networks exploiting both magnitude and phase data in SAR under National Geospatial-Intelligence Agency (NGA) Boosting Innovative GEOINT BAA.
• Evaluated domain shift bias from training with synthetic data; combined collected with synthetic data using transfer learning to significantly improve model performance.
• Redeveloped existing models to modern SOTA single-shot architectures improving performance.
• Presented work at 2020 SAREM forum in Chantilly, Virginia.

Machine learning and autonomy, worked with the Cognitive Autonomy Research Group.

• Developed and trained deep learning models for multiple projects including perception with satellite imagery and anomaly detection with time-series telemetry data.
• Developed trajectory optimization and nonlinear state estimation algorithms as well as accompanying simulation platform for autonomous formation flying.

Machine learning, Guidance, Navigation & Control (GNC), and modeling/simulation for national security space in support of the U.S. Space Force and National Reconnaissance Office (NRO).

• Lead a team working on the next generation GPS satellite constellation (GPS Block III) and worked on other satellites, launch vehicles (Atlas V and Falcon), and missile systems.

Guidance, Navigation & Control (GNC) and modeling/simulation for satellites.

• Developed new control layer capable of autonomously navigating satellites through on-orbit scenarios from the ground and software tools to improve autonomous accounting of propellant usage.

Guidance, Navigation & Control (GNC) and modeling/simulation for the GOES satellite constellation operated by NOAA.

• Presented a paper on thruster leak analysis at the SpaceOps 2010 Conference in Huntsville, Alabama.