Skip to content
@SpeyTech

SpeyTech

Deterministic computing for safety-critical systems

SpeyTech

Deterministic computing for safety-critical systems.

What is this?

Open source tools and libraries for building systems where reproducibility and correctness are non-negotiable. Fixed-point arithmetic, deterministic ML inference, and safety kernel patterns — all designed to support certification under DO-178C, ISO 26262, and IEC 62304.

Why does it exist?

Floating-point non-determinism, hidden race conditions, and opaque ML models don't belong in systems that control aircraft, medical devices, or vehicles. These projects demonstrate that deterministic alternatives exist — and can be practical.

Projects

Certifiable ML Pipeline

Education

Tools

How do I use it?

Each repository includes documentation in docs/ with requirements, build instructions, and examples. Clone, build, run the tests.

How do I contribute?

See CONTRIBUTING.md. Issues and pull requests welcome.

Learn more

speytech.com


Copyright © 2026 The Murray Family Innovation Trust. All rights reserved.

Pinned Loading

  1. c-from-scratch c-from-scratch Public

    Learn to build safety-critical systems in C. Prove first, code second.

    C 290 24

  2. c-sentinel c-sentinel Public

    Semantic Observability for UNIX Systems - A lightweight C-based system prober with AI-powered analysis

    C 71 7

  3. certifiable-inference certifiable-inference Public

    Deterministic, bit-perfect AI inference for safety-critical systems

    C 2 3

  4. certifiable-bench certifiable-bench Public

    Performance benchmarking for deterministic ML inference. Measures latency, throughput, WCET with cryptographic verification of bit-identity across platforms. 11,840 test assertions. Pure C99, no dy…

    C

  5. certifiable-training certifiable-training Public

    Deterministic ML training for safety-critical systems. Fixed-point arithmetic, Merkle audit trails, zero dynamic allocation. DO-178C / IEC 62304 / ISO 26262 ready

    C 1

  6. fixed-point-fundamentals fixed-point-fundamentals Public

    Educational course teaching fixed-point arithmetic from first principles. Strict C99, no dependencies, MIT licensed.

    C 1 1

Repositories

Showing 10 of 14 repositories
  • c-from-scratch Public

    Learn to build safety-critical systems in C. Prove first, code second.

    SpeyTech/c-from-scratch’s past year of commit activity
    C 290 MIT 24 0 0 Updated Jan 24, 2026
  • certifiable-verify Public

    Pipeline verification for certifiable-* ecosystem. Validates complete provenance chain from data through deployment with cryptographic binding verification. Hash-only and full replay modes.

    SpeyTech/certifiable-verify’s past year of commit activity
    C 0 GPL-3.0 0 4 0 Updated Jan 23, 2026
  • certifiable-training Public

    Deterministic ML training for safety-critical systems. Fixed-point arithmetic, Merkle audit trails, zero dynamic allocation. DO-178C / IEC 62304 / ISO 26262 ready

    SpeyTech/certifiable-training’s past year of commit activity
    C 0 GPL-3.0 1 4 0 Updated Jan 23, 2026
  • certifiable-quant Public

    Deterministic, certifiable model quantization (FP32→Q16.16) with formal error bounds and cryptographic proof certificates. Pure C99. Zero malloc. DO-178C/IEC 62304/ISO 26262 ready.

    SpeyTech/certifiable-quant’s past year of commit activity
    C 0 GPL-3.0 1 4 0 Updated Jan 23, 2026
  • certifiable-monitor Public

    Deterministic runtime monitoring for safety-critical ML. Hash-chained audit ledger, drift detection (TV/JSD/PSI), COE policy enforcement, and tamper-evident logging. 253 tests passing.

    SpeyTech/certifiable-monitor’s past year of commit activity
    C 0 GPL-3.0 0 4 0 Updated Jan 23, 2026
  • certifiable-inference Public

    Deterministic, bit-perfect AI inference for safety-critical systems

    SpeyTech/certifiable-inference’s past year of commit activity
    C 2 GPL-3.0 3 4 0 Updated Jan 23, 2026
  • certifiable-harness Public

    End-to-end test harness proving bit-identity across platforms. Runs all 7 certifiable-* stages, generates golden references, and verifies deterministic execution. Proven identical: Linux/GCC ↔ macOS/Clang.

    SpeyTech/certifiable-harness’s past year of commit activity
    C 0 GPL-3.0 0 4 0 Updated Jan 23, 2026
  • certifiable-deploy Public

    Deterministic model packaging and cryptographic attestation for safety-critical ML deployment

    SpeyTech/certifiable-deploy’s past year of commit activity
    C 0 GPL-3.0 0 4 0 Updated Jan 23, 2026
  • certifiable-data Public

    Deterministic data pipeline for safety-critical ML systems. Fixed-point arithmetic, Feistel shuffling, Merkle audit trails. Zero dynamic allocation. Certifiable for DO-178C, IEC 62304, ISO 26262. 142 tests passing.

    SpeyTech/certifiable-data’s past year of commit activity
    C 0 GPL-3.0 0 4 0 Updated Jan 23, 2026
  • certifiable-bench Public

    Performance benchmarking for deterministic ML inference. Measures latency, throughput, WCET with cryptographic verification of bit-identity across platforms. 11,840 test assertions. Pure C99, no dynamic allocation. Part of the certifiable-* ecosystem for safety-critical AI.

    SpeyTech/certifiable-bench’s past year of commit activity
    C 0 GPL-3.0 0 0 0 Updated Jan 23, 2026

People

This organization has no public members. You must be a member to see who’s a part of this organization.

Top languages

C