AUSTIN, TX – February 25, 2025 – Imandra Inc., a neurosymbolic reasoning AI startup, today announced ImandraX, the latest release of its Imandra automated reasoning engine.
The company said ImandraX marks an advancement in AI-driven logical reasoning, introducing powerful reasoning algorithms and architectural features for integration with AI agents.
With the increasing reliance on AI in high-stakes industries—including finance, healthcare, and autonomous systems—trustworthy, explainable, and mathematically rigorous reasoning is more critical than ever. ImandraX sets a new standard in AI-driven logical analysis with cutting-edge innovations in proof automation, counterexample generation, and decision procedures that improve both performance and reliability.
“ImandraX is a transformational step in making advanced symbolic reasoning a core part of AI workflows,” said Denis Ignatovich, Co-founder of Imandra Inc. “By equipping AI agents with powerful automated logical reasoning and formal verification capabilities, we’re pushing the boundaries of what intelligent systems can achieve.”
Co-founder Dr Grant Passmore added: “ImandraX is the culmination of years of research and real-world deployment across some of the most demanding industries—including finance, defense, and AI. Our customers and partners rely on Imandra’s automated reasoning to ensure the safety and reliability of mission-critical systems, from financial exchanges to autonomous agents. With ImandraX, we’re not just making rigorous reasoning accessible—we’re making it indispensable for the next generation of AI-powered decision-making.”
ImandraX introduces a range of capabilities, including:
- Breakthroughs in Proof Automation – Introduces new techniques for mixed discrete and continuous recursive functions over datatypes, enabling the first formal model and verification of the new IEEE P3109 standard for small (<16) bit binary floating-point formats. These new formats are essential for accurate quantization and distillation of neural networks and are being developed by the IEEE P3109 working group.
- Progress in Neural Network Safety Verification –Integrates higher-order bounded model checking and induction, powering the first formally verified proof checker for neural network safety property verification.
- Foundational Advancements in State-Space Region Decomposition – Delivers >4x speedups for region decomposition jobs, significantly improving efficiency for finance users generating high-coverage test suites for FIX connectivity and other mission-critical applications.
- Enhanced Developer Experience – A new VS Code plugin enables parallel proof development with concurrent jobs running in Imandra’s reasoning cloud, streamlining the formal verification workflow.
