Keep pulling the thread on Cameron D'Ambrosi and Jason Hoch.
SpaceX made an early strategic decision to build its own proprietary, in-house software for test data analysis because existing commercial solutions were deemed inadequate.
Nominal plans to begin building its own hardware sooner rather than later.
Nominal's platform is used by four of the top five defense prime contractors in the United States.
Nominal's platform helps robotics companies improve their model training data by identifying and filtering out data segments where physical components were malfunctioning or out of calibration.
Nominal is working with the U.S. Air Force and DARPA on a project called CIFR (Cyber Physical Systems Executing in Real Time).
The DARPA project CIFR involves using AI agents paired with digital twins to operate faster than real-time and recommend the next optimal, knowledge-maximizing test condition for a physical system.
The CIFR project aims to shift hardware testing from a sequential, deterministic matrix of test points to a continuous, gradient-based approach where an AI agent rapidly adjusts test parameters in real-time.
Nominal's strategy includes building a large, proprietary dataset of human-enriched hardware data, created by mechanical engineers using its platform.
Unlike software development where testing is a largely solved problem, hardware engineering for software-defined hardware, autonomy, and robotics lacks mature testing solutions.
Companies like SpaceX, Anduril, and Tesla have demonstrated that investing in internal infrastructure and tools for hardware development provides a significant competitive advantage.
Based on his experience at Anduril, Cameron D'Ambrosi believes the most effective approach for hardware development is to blend simulation outputs with real-world telemetry and sensor data.
The proprietary simulation technologies developed over decades at traditional defense prime contractors are being rapidly disrupted by newer companies like Anduril.