# 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 Definitions

### What is deterministic computing?
Deterministic computing produces identical outputs for identical inputs across every run, platform, and build by eliminating non-deterministic execution paths.

### What is deterministic inference?
Deterministic inference produces bit-identical neural network outputs across platforms and builds by using fixed-point arithmetic and a single canonicalised execution path.

### What is fixed-point arithmetic?
Fixed-point arithmetic represents real numbers as scaled integers, eliminating IEEE 754 floating-point variance and ensuring identical results across platforms under C99.

### What is a certifiable ML pipeline?
A certifiable ML pipeline produces bit-identical, auditable outputs at every stage — data, training, quantisation, inference — with cryptographic proof of each transition.

### What is the MDCP (Murray Deterministic Computing Platform)?
MDCP is a tick-based deterministic execution substrate providing mathematical reproducibility guarantees for safety-critical systems through architectural constraints. Patent GB2521625.0.

### What is canonicalisation in deterministic systems?
Canonicalisation transforms data into a single, unambiguous byte representation so that identical inputs always produce identical hashes regardless of struct layout or compiler.

### Why does floating-point arithmetic break reproducibility?
IEEE 754 floating-point permits platform-specific optimisations such as FMA fusion, producing different results on x86 vs ARM and making bit-identical reproducibility impossible.

### What is safety-critical software certification?
Safety-critical certification (DO-178C, IEC 62304, ISO 26262) requires provable, reproducible behaviour — deterministic systems satisfy this where conventional systems cannot.

### What is verifiable AI execution?
Verifiable AI execution provides cryptographic proof that an AI system ran correctly — same inputs, same model, same outputs — with an auditable trail of every decision.

### What is round-to-nearest-even (RNE) rounding?
Round-to-nearest-even rounds 0.5 cases to the nearest even digit, eliminating systematic bias and ensuring bit-identical results across platforms for fixed-point ML.

## 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/SpeyTech/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/SpeyTech/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/SpeyTech/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/SpeyTech/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/SpeyTech/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/SpeyTech/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/SpeyTech/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/SpeyTech/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/SpeyTech/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/SpeyTech/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 (40 articles)
- [AI Architecture](https://speytech.com/ai-architecture/) - Production AI systems, MLOps patterns, and certifiable ML (17 articles)
- [Open Source](https://speytech.com/open-source/) - Documentation for open source projects (12 projects)

## Recent Insights Articles

- [One IP, Six Crawler Identities, One Second: Detection Built Against Real Production Logs](https://speytech.com/insights/rotational-bot-identity-detection/) - How do you detect a single attacker IP claiming multiple trusted crawler identities while probing security-sensitive paths in nginx access logs, and what failure modes does the detector itself ship with that only production observation surfaces after the synthetic test suite has already passed?
  - Defines: Rotational bot-identity spoofing
- [SEO Validator: A Deploy Gate for SEO Regressions](https://speytech.com/insights/seo-validator-deploy-gate/) - How can SEO regression testing be integrated as a deploy gate that automatically rolls back broken builds, and why does an SEO validator that reads nginx configuration files catch redirect and asset failures that traditional URL-list-based SEO audit tools miss?
  - Defines: Deploy-gate SEO validation
- [Self-Hosted Static Sites Need Operational SEO Observability](https://speytech.com/insights/operational-seo-observability/) - How do you detect crawler-visible SEO failures on self-hosted static sites when conventional tools like Search Console and Lighthouse cannot see them, and what does the operational recovery actually look like in measured nginx log data when sitemap freshness signals are remediated and the site's Googlebot crawl rate jumps from a 6.83-per-day baseline?
  - Defines: Operational SEO observability
- [Undefined Behaviour: Compiler's Licence to Delete Your Code](https://speytech.com/insights/undefined-behaviour-compiler/) - What is undefined behaviour in C, how do C compilers exploit undefined behaviour to remove safety checks and reorder code, what are the most common sources of undefined behaviour including signed overflow and null pointer dereference and strict aliasing, and how do compiler flags like -fwrapv and -fno-strict-aliasing control UB-related optimisations?
  - Defines: Undefined behaviour
- [Why memcmp Fails on Structs: Padding, Floats, Silent Bugs](https://speytech.com/insights/memcmp-struct-comparison/) - Why does memcmp produce incorrect results when comparing C structs containing padding bytes or floating-point fields, how do positive zero and negative zero and NaN break byte-level equality, and what is the correct approach to struct comparison in safety-critical C code?
  - Defines: Struct byte comparison
- [The Type Promotion Trap: C's Silent Integer Conversion Bugs](https://speytech.com/insights/type-promotion-trap/) - How do C's integer promotion rules cause silent bugs when comparing signed and unsigned integers, why does the expression 0u > -1 evaluate to false, what are the common type conversion patterns that create logic errors in embedded and safety-critical C code, and which compiler flags like -Wsign-compare and -Wconversion catch implicit promotion bugs?
  - Defines: Integer promotion
- [When Fixed-Point Beats Floating-Point (And When It Doesn't)](https://speytech.com/insights/fixed-point-vs-floating-point-tradeoffs/) - When is Q16.16 fixed-point arithmetic the right engineering choice over IEEE 754 floating-point for ML inference in safety-critical systems, when is floating-point perfectly acceptable, and what are the honest trade-offs in numerical precision, dynamic range, runtime performance, and certification effort?
- [Init-Update-Status-Reset: O(1) Safety Guarantees](https://speytech.com/insights/init-update-status-reset-pattern/) - What is the Init-Update-Status-Reset four-function interface pattern used in safety-critical C programming, and how does this fixed and minimal API contract enable O(1) resource bounds, static analysis feasibility, bounded deterministic memory usage, and compositional verification across all modular system components?
  - Defines: Init-Update-Status-Reset
- [Contracts as Documentation: Why Comments Lie](https://speytech.com/insights/contracts-as-documentation/) - How do software contracts using explicit preconditions, postconditions, and class invariants serve as machine-verifiable living documentation that stays permanently synchronised with actual source code behaviour, and why are traditional code comments and separate external documents unreliable compared to enforced contract specifications?
  - Defines: Software contracts
- [Why 'Hello World' Fails Safety-Critical Engineers](https://speytech.com/insights/hello-world-fails-safety-critical/) - Why does the traditional C hello world programme and conventional C tutorial teaching approach actively fail to prepare engineers for safety-critical software development, and what specific coding habits and assumptions do certification auditors routinely reject that introductory tutorials commonly teach?
  - Defines: Safety-critical C

## Recent AI Architecture Articles

- [Incident Reconstruction: Beyond It Worked Yesterday](https://speytech.com/ai-architecture/incident-reconstruction-determinism/) - How does deterministic execution architecture with bit-perfect replay capability, cryptographic execution tracing through hash chains, and tamper-evident sealed audit logs transform incident reconstruction and root cause analysis in safety-critical systems from conventional log-based guesswork into rigorous and repeatable computational forensics?
  - Defines: Deterministic incident reconstruction
- [Version Control for Deterministic Systems: Git Isn't Enough](https://speytech.com/ai-architecture/version-control-deterministic-systems/) - Why is Git alone insufficient for version control and configuration management in deterministic safety-critical systems requiring formal certification evidence, and how do Merkle chains, cryptographic attestation of build artifacts, and fully reproducible builds provide the tamper-evident evidence trail that standards demand?
  - Defines: Merkle chain version control
- [Testing ML Systems: Beyond Unit Tests and Accuracy Metrics](https://speytech.com/ai-architecture/ml-testing-systems/) - What testing strategies go beyond basic unit tests and aggregate accuracy metrics for production machine learning systems, and how do you systematically test for data distribution drift, model degradation over time, infrastructure integration failures, and edge case behaviour in ML pipelines?
- [Cost Engineering for ML Infrastructure](https://speytech.com/ai-architecture/ml-cost-engineering/) - Where does the money actually go in production machine learning infrastructure, which cost categories do engineering teams most commonly misjudge or overlook, and what should teams optimise first when managing compute costs across model training, real-time inference, data storage, and pipeline operations?
- [State Management in ML Services: Beyond Stateless Inference](https://speytech.com/ai-architecture/ml-state-management/) - What architectural patterns handle state management in ML inference services that need to maintain request context, session history, or feature cache state between sequential requests, and when should engineering teams move beyond simple stateless inference to implement stateful model serving patterns?
- [Graceful Degradation in ML Systems](https://speytech.com/ai-architecture/ml-graceful-degradation/) - What are the proven architectural patterns for implementing graceful degradation in production ML inference systems, and how do you design reliable fallback strategies for common failure scenarios when models time out, feature stores become unreachable, memory pressure spikes, or GPU hardware errors occur?
- [The Observability Blind Spot: What ML Metrics Miss](https://speytech.com/ai-architecture/ml-observability-blind-spot/) - What critical information about production model health do standard ML monitoring metrics like accuracy, latency, and throughput fail to capture, and how do these observability blind spots allow silent model failures and data drift to go entirely undetected by conventional dashboards?
- [The Certifiable-* Ecosystem: One Deterministic ML Pipeline](https://speytech.com/ai-architecture/certifiable-ecosystem/) - What is the certifiable-* ecosystem of eight interconnected C99 projects and how does it provide a complete deterministic ML pipeline with cryptographic provenance chains, fixed-point arithmetic, and static allocation for safety-critical systems that require bit-identical execution across different platforms and compilers?
  - Defines: Certifiable-* ecosystem
- [Deterministic ML Pipeline for Safety-Critical Systems](https://speytech.com/ai-architecture/deterministic-ml-pipeline/) - How do you build a complete deterministic machine learning pipeline for safety-critical systems that guarantees bit-identical output across every platform, covering each stage from data loading through training, quantisation, deployment packaging, and inference using fixed-point arithmetic and cryptographic audit trails?
  - Defines: Deterministic ML pipeline
- [WCET Analysis for Neural Network Inference](https://speytech.com/ai-architecture/wcet-neural-network-inference/) - How do you perform worst-case execution time analysis for neural network inference operations including convolution, matrix multiplication, pooling, and activation functions, and why is provable WCET boundedness through static timing analysis essential for certifying ML inference in hard real-time safety-critical systems?
  - Defines: WCET analysis

## Open Source Project Documentation

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

## 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/