# SpeyTech

> Deterministic computing systems for safety-critical environments.

SpeyTech develops deterministic software platforms for aerospace, medical devices, and autonomous systems. Founded by William Murray, a Regenerative Systems Architect with 30 years of UNIX infrastructure experience, based in the Scottish Highlands.

## Core Products

- [MDCP](https://speytech.com/mdcp/) - Murray Deterministic Computing Platform. Tick-based deterministic execution substrate for safety-critical systems. Patent GB2521625.0.
- [MDLCE](https://speytech.com/mdlce/) - Murray Deterministic Liability Closure Engine. Cryptographic execution binding for compliance attestation and post-incident analysis. Patent GB2522369.4.
- [CardioCore](https://speytech.com/cardiocore/) - Deterministic medical device kernel for implantable devices. Aligned with IEC 62304 Class C requirements.

## Open Source Projects

- [Fixed-Point Fundamentals](https://github.com/SpeyTech/fixed-point-fundamentals) - Learn fixed-point arithmetic from first principles — because 'close enough' isn't deterministic MIT licensed. ([docs](https://speytech.com/open-source/fixed-point-fundamentals/))
- [certifiable-bench](https://github.com/SpeyTech/certifiable-bench) - Performance benchmarking for deterministic ML — because 'fast' means nothing if you can't prove it's correct GPL-3.0 licensed. ([docs](https://speytech.com/open-source/certifiable-bench/))
- [certifiable-harness](https://github.com/williamofai/certifiable-harness) - End-to-end test harness for deterministic ML — because 'it works on my machine' isn't certifiable GPL-3.0 licensed. ([docs](https://speytech.com/open-source/certifiable-harness/))
- [certifiable-verify](https://github.com/williamofai/certifiable-verify) - Pipeline verification for the certifiable-* ecosystem — because 'we checked it manually' isn't certifiable GPL-3.0 licensed. ([docs](https://speytech.com/open-source/certifiable-verify/))
- [certifiable-monitor](https://github.com/williamofai/certifiable-monitor) - Deterministic runtime monitoring — because 'the model drifted' isn't certifiable GPL-3.0 licensed. ([docs](https://speytech.com/open-source/certifiable-monitor/))
- [Certifiable Deploy](https://github.com/williamofai/certifiable-deploy) - Deterministic model packaging and cryptographic attestation — because 'trust me, it's the right model' isn't certifiable GPL-3.0 licensed. ([docs](https://speytech.com/open-source/certifiable-deploy/))
- [Certifiable Quant](https://github.com/williamofai/certifiable-quant) - Deterministic model quantization with formal error certificates for safety-critical ML GPL-3.0 licensed. ([docs](https://speytech.com/open-source/certifiable-quant/))
- [C-From-Scratch](https://github.com/williamofai/c-from-scratch) - Learn to build safety-critical systems in C — mathematical rigour, not 'Hello World' MIT licensed. ([docs](https://speytech.com/open-source/c-from-scratch/))
- [Certifiable Data](https://github.com/williamofai/certifiable-data) - Deterministic data pipelines for safety-critical ML — because 'we shuffled the data' isn't reproducible GPL-3.0 licensed. ([docs](https://speytech.com/open-source/certifiable-data/))
- [C-Sentinel](https://github.com/williamofai/c-sentinel) - Semantic observability for UNIX systems — lightweight system probing with explainable risk scoring MIT licensed. ([docs](https://speytech.com/open-source/c-sentinel/))
- [Certifiable Inference](https://github.com/williamofai/certifiable-inference) - Deterministic, bit-perfect neural network inference for safety-critical systems GPL-3.0 licensed. ([docs](https://speytech.com/open-source/certifiable-inference/))
- [Certifiable Training](https://github.com/williamofai/certifiable-training) - Deterministic ML training with Merkle audit trails — because 'we trained it' isn't certifiable GPL-3.0 licensed. ([docs](https://speytech.com/open-source/certifiable-training/))

## Key Topics

- Deterministic computing and tick-based scheduling
- Safety-critical certification (DO-178C, IEC 62304, ISO 26262)
- Fixed-point arithmetic and Q notation
- Bit-identical cross-platform execution
- Certifiable machine learning
- Formal verification and mathematical proofs
- Cryptographic execution tracing
- Round-to-nearest-even (RNE) rounding
- Merkle audit trails for ML training

## Content Sections

- [Why Deterministic Computing](https://speytech.com/why-deterministic-computing/) - Introduction to the value proposition
- [Insights](https://speytech.com/insights/) - Technical articles on deterministic systems, formal methods, and safety certification (29 articles)
- [AI Architecture](https://speytech.com/ai-architecture/) - Production AI systems, MLOps patterns, and certifiable ML (10 articles)
- [Open Source](https://speytech.com/open-source/) - Documentation for open source projects (12 projects)

## Recent Insights Articles

- [Stochastic Rounding Without the Stochastic](https://speytech.com/insights/stochastic-rounding-deterministic/) - How PRNG-controlled rounding can provide regularisation benefits deterministically
- [Cross-Platform Bit-Identity: From Theory to 7 Matching Hashes](https://speytech.com/insights/cross-platform-bit-identity/) - The practical journey of verifying deterministic ML across platforms
- [The Feistel Shuffle: Deterministic Data Ordering Without Randomness](https://speytech.com/insights/feistel-shuffle-deterministic/) - How cycle-walking Feistel networks can provide reproducible shuffling for ML training
- [Merkle Chains for ML Audit Trails](https://speytech.com/insights/merkle-chains-ml-audit/) - How cryptographic hash chains can make every training step verifiable
- [Round-to-Nearest-Even: The Rounding Mode That Makes Determinism Possible](https://speytech.com/insights/round-to-nearest-even/) - Why banker's rounding matters for bit-identical machine learning
- [Fixed-Point Neural Networks: The Math Behind Q16.16](https://speytech.com/insights/fixed-point-neural-networks/) - How integer arithmetic can enable deterministic AI inference for safety-critical systems
- [Bit-Perfect Reproducibility: Why It Matters and How to Prove It](https://speytech.com/insights/bit-perfect-reproducibility/) - What deterministic execution actually means and how to verify it across platforms
- [The Real Cost of Dynamic Memory in Safety-Critical Systems](https://speytech.com/insights/dynamic-memory-safety-critical/) - Why malloc is problematic for certification and how static allocation can simplify verification
- [From Proofs to Code: Mathematical Transcription in C](https://speytech.com/insights/mathematical-proofs-to-code/) - How mathematical contracts become deterministic implementations
- [The Hidden Cost of Non-Determinism](https://speytech.com/insights/nondeterminism-cost-debugging/) - Understanding the financial impact of debugging race conditions and Heisenbugs

## Recent AI Architecture Articles

- [The Certifiable-* Ecosystem: Eight Projects, One Deterministic ML Pipeline](https://speytech.com/ai-architecture/certifiable-ecosystem/) - From training data to deployed inference — bit-identical, auditable, certifiable
- [A Complete Deterministic ML Pipeline for Safety-Critical Systems](https://speytech.com/ai-architecture/deterministic-ml-pipeline/) - From training data to deployed inference — bit-identical, auditable, certifiable
- [WCET Analysis for Neural Network Inference](https://speytech.com/ai-architecture/wcet-neural-network-inference/) - How to prove worst-case execution time for convolution, matrix multiply, and pooling operations
- [Why TensorFlow Lite Faces Challenges in DO-178C Certification](https://speytech.com/ai-architecture/tflite-do178c-challenges/) - Understanding the architectural properties that complicate aerospace certification for mobile inference frameworks
- [Why Floating Point Is Dangerous: The Case for Deterministic AI in C](https://speytech.com/ai-architecture/floating-point-danger/) - When 'mostly reproducible' isn't good enough for systems that matter
- [Debugging Model Behavior in Production](https://speytech.com/ai-architecture/debugging-model-behavior-production/) - When the model works in staging but fails in prod, here's how to find out why
- [When You Don't Need a Feature Store](https://speytech.com/ai-architecture/when-you-dont-need-feature-store/) - Most teams solve a problem they don't have yet
- [Model Serving Architecture Patterns](https://speytech.com/ai-architecture/ai-model-serving-patterns/) - Understanding latency, throughput, and the trade-offs between them
- [Production AI Systems: What 30 Years of UNIX Taught Me](https://speytech.com/ai-architecture/production-ai-unix-principles/) - The infrastructure principles that kept systems running still apply to ML
- [The Observability Gap in ML Systems](https://speytech.com/ai-architecture/ml-observability-gap/) - Why your model serving cluster fails at 3AM and you can't figure out why

## Technical Demonstrations

- [Tick Scheduler](https://speytech.com/tick-scheduler/) - Interactive demonstration of deterministic scheduling
- [Incident Replay](https://speytech.com/replay-demo/) - Byte-identical execution replay visualisation
- [Race Conditions](https://speytech.com/race-visualizer/) - Why conventional systems fail under concurrency
- [ROI Calculator](https://speytech.com/asil-calculator/) - Certification cost reduction estimates

## About the Founder

William Murray is a Visiting Scholar at Heriot-Watt University with 30 years of UNIX systems engineering experience. Sole inventor of MDCP (Patent GB2521625.0) and MDLCE (Patent GB2522369.4). Based in Inverness, Scottish Highlands, UK.

## Contact

- Website: https://speytech.com/contact/
- GitHub: https://github.com/SpeyTech
- LinkedIn: https://www.linkedin.com/in/william-murray-5180aa32b/