Avometric LLC
Avometric pairs AI-assisted engineering with deterministic, mathematically rigorous verification to deliver systems that are provably correct in the domains where failure is not an option: embedded, safety-critical, and national security missions.
Avometric is an R&D-seeded small business developing a portfolio of solutions for national security and assured autonomy missions. We are led by senior researchers who have repeatedly transitioned government research into commercial products, open-source libraries, and programs of record.
Our work centers on high-assurance software and security: formal verification of real code, AI-assisted reverse engineering and attack surface analysis, verifiable and resilient computation, and quantum-secure systems. We pair AI-driven analysis with deterministic verification oracles so that findings are not just plausible but provable.
We are an active, cleared performer on advanced government R&D, including ongoing DARPA work, across the Department of War and the broader defense and intelligence community. The same core IP that proves out on today's programs is the foundation for a commercial assured-autonomy capability over the longer term.
A focused set of deep technical competencies, composed into end-to-end assured-autonomy pipelines.
Machine-checked proofs over real source code, covering memory safety, functional bounds, and constant-time properties, using Frama-C/ACSL, Dafny, and Lean 4. Demonstrated on production cryptographic and networking code, with a certification pathway to DO-178C and DO-333.
A four-layer AI-augmented analysis pipeline, spanning source and binary static analysis, instrumented emulation, and structured reporting, that generates and refines its own tooling. Built for the sparse, proprietary, hardware-coupled targets that autonomous commodity tools do not cover.
Contract-driven, human-on-the-loop agentic systems for defensive cyber operations and security assessment. A machine-readable mission contract bounds what the agent may do, a deterministic oracle checks every action, and a human keeps authority over consequential steps, so autonomy stays inside policy.
Adversarial review that finds the specification gaps, emergent multi-layer threat chains, and side-channel exposure that formal methods cannot see by construction. We run formal verification and attack-surface analysis together and show where each one is load-bearing.
Runtime fault detection, correction, and computational-integrity assurance for untrusted and opaque processors, including the first known practical runtime error-correcting computation, delivered under a DARPA program.
Quantum-resistant hardening of fielded systems: integration, optimization, and evaluation of the NIST-standardized quantum-secure algorithms ML-KEM and ML-DSA on constrained and FPGA-accelerated targets, delivered with machine-checked constant-time proofs and a clear CNSA 2.0 transition path.
Adversary-tolerant state estimation, sensor fusion, and probabilistic modeling for contested and degraded environments, built to hold up against deception, jamming, and unreliable or manipulated data, and applied to problems such as supply-chain and mission risk.
Secure firmware and capability upgrades for size-, weight-, power-, and cost-constrained edge platforms, including radiation-hardened and certification-frozen processors. We add new, formally assured functionality to already-fielded hardware without new silicon, board redesign, or recertification, offloading heavy computation to spare FPGA and DSP fabric.
Independent verification and adversarial test and evaluation for assured autonomy and security-critical systems, the evaluator and integrator role we held on a DARPA program, extended toward assurance for decentralized, multi-agent autonomy.
Co-founder
Former Partner and Applied Research Manager in AI for Cybersecurity at Microsoft Corporation; VP of Advanced Cyber Science at CACI Federal (formerly LGS Innovations); former leader of DARPA cyber programs. PhD and ScM in Computer Science, Brown University.
Co-founder
CTO and Co-founder of Duality Technologies, a VC-backed startup commercializing DARPA-funded research; creator of the OpenFHE open-source library and principal investigator on DARPA cryptography programs. Associate Professor and DARPA Director's Fellow, NJIT. PhD and Master of Electrical Engineering, University of Michigan.
We work with government program offices across DARPA and the broader Department of War, defense primes, design partners, and early investors exploring assured autonomy for mission-critical systems.