< Services />
I'm the person you bring in when your system works most of the time, and nobody can tell you why it sometimes doesn't. Every engagement is scoped to a concrete outcome — not billed hours against a vague brief.
🌱 Starter Reliability Review · founding client
A fixed-price, read-only first look for small businesses and founders — the one small service or pipeline most likely to fail quietly. Nothing is touched or changed.
- One small service or pipeline — a slice, not a whole estate.
- A read-only pass over its config and structure.
- A short, prioritized findings note in plain language.
- One follow-up email on the findings.
🔍 Reliability Review · read-only
Your system works most of the time and no one can say why it sometimes doesn't. I read the detail: your configuration, pipelines, and validation, and report exactly where a wrong value or a missing guardrail will fail silently. You get prioritized findings plus the guardrail specs for your own engineers to implement. Nothing in your system is touched or changed.
📐 Architecture Review
The failure modes, trust boundaries, and data flow that hold your system together live in people's heads instead of on paper, so no one can point to where the design itself is fragile. I map the structure: how the pieces connect, what trusts what, where data flows, and hand back a prioritized read of the weak points in the design and what to rethink first.
🚧 Validation & Guardrails · hands-on
Everything in the Reliability Review, plus the fix: where the read-only pass finds the gaps, I build the guardrails that close them. Schema-driven validation (JSON Schema, YAML, typing.Annotated) that rejects invalid configuration and data with a clear, human-readable error before your system ever runs — a preflight check that stops the bad state at the door, not after it ships. I work in a branch or environment you control, and confirm the guardrails catch known-bad configurations before handover. Nothing is attacked, and your production environment is never touched.
🛠️ Developer Tooling & CLIs
Your team loses hours to fragile scripts, manual steps, and tools only one person really understands. I build internal tools, CLIs, and developer-experience improvements that turn that friction into something anyone can run without holding their breath.
🔬 Scientific Software
Your research pipeline gives a slightly different answer each time it runs, and collaborators or reviewers can't reproduce your results — the kind of thing that stalls papers and quietly erodes trust in the data. I build Python tooling and configuration systems that make simulation and research pipelines reproducible — the same work shipped in the official ESA Pyxel 2.14 release.
Trusted Partner
Justice Akason
I have partnered with Justice Akason and CrenavoLab for cybersecurity expertise on projects and products where deeper security assessment, hardening, and security-focused engineering are required.
Justice and the CrenavoLab platform provide the cybersecurity perspective for Nimbus Vestige, my self-hosted forensic reconstruction platform for cloud and identity incidents. Through this partnership, we combine software engineering and reliability expertise with specialized cybersecurity knowledge to build more secure and resilient products.
Visit CrenavoLab →
< LLM Workflow Router />
Reliable Infrastructure for AI Workflows
The LLM Workflow Router helps engineering teams build AI applications that are predictable, observable, and easier to operate in production. It validates workflow topology before execution, eliminates invalid routing paths, exposes explicit failure states, and integrates with OpenTelemetry for end-to-end visibility—reducing the risk of unexpected behaviour in complex AI systems.
Inspired by the Lumenoid AI Framework, the router focuses on structural safety rather than content moderation. Lumenoid helps applications make responsible AI decisions, while the LLM Workflow Router ensures those decisions flow through explicit, deterministic, and auditable workflows.
Source-Available Core • Commercial License for Production Use
pip install llm-workflow-routerFree to inspect, run, and evaluate — always. Read the full source, try it against your own data, and satisfy yourself it's right. A commercial license is required for any commercial or production use, and includes trust-boundary enforcement, policy validation, tamper-evident auditing, and support.
Read every line before you pay. The full source is open for inspection — nothing hidden, nothing to take on faith.
🛰️ Inspired by the Lumenoid AI Framework
Lumenoid is an open ethical AI architecture that defines reference principles for responsibility, transparency, uncertainty, bounded capability, and human-centred system design. Rather than prescribing model behaviour, it provides architectural foundations that help AI systems remain understandable, accountable, and safe to evolve.
🛡️ Licensing
The core LLM Workflow Router is source-available under the PolyForm Noncommercial License 1.0.0 — free to read, run, and evaluate, with commercial and production use reserved for licensees.
A commercial license grants production use and includes
enterprise-grade trust boundaries, policy enforcement,
tamper-evident audit trails, and the
wfrouter.security.* telemetry convention.
Commercial subscriptions include ongoing security updates, evolving policy packs, compatibility with OpenTelemetry GenAI conventions, and priority support.
💼 Commercial Security Licensing
Early access. The Security Layer is onboarding a small founding cohort. Founding licenses come at introductory pricing in exchange for feedback, and lock in that rate as the product matures. Request current terms and I'll send the details.
| Tier | Founding Access | Which one am I? |
|---|---|---|
| Indie | from £290/yr · founding | I'm shipping it in my own product, solo |
| Team | from £1,200/yr · founding | My team runs it in production |
| Enterprise | Let's talk | Company-wide, or we resell / embed it |
< Nimbus Vestige />
Forensic Reconstruction for Cloud & Identity Incidents
Nimbus Vestige helps incident responders answer the question detection tools can't: how did the intrusion actually happen? Given fragmented control-plane, sign-in, and consent telemetry, it reconstructs the most likely path an attacker took across accounts and services, ranks the alternative paths they could have taken, and reports every step with calibrated, evidence-traceable confidence.
Honesty is the product. Every number is traceable to the specific evidence that earned it — and anything the data cannot support is withheld rather than guessed. The engine is substrate-independent: the same abductive core that reconstructs an Entra ID / M365 intrusion reconstructs an AWS one. And because NV runs entirely self-hosted in your environment, your incident telemetry never leaves your infrastructure.
Source-Available Core • Self-Hosted • Commercial License for Production Use
git clone https://gitlab.com/dobybaxter127/nimbus-vestigeFree to inspect, run, and evaluate — always. Clone the full source, try it against your own data on the live demo, and satisfy yourself it's right. A commercial license is required for any commercial or production use, and includes the auto-updating pattern feeds, calibration tooling, and priority support for real investigations — pay once, keep it for good.
Read every line before you pay. The full source is open for inspection — nothing hidden, nothing to take on faith.
🛡️ Licensing
The Nimbus Vestige core is source-available under the PolyForm Noncommercial License 1.0.0 — the full reconstruction engine, both provider ingests, GUI, and offline eval harness, free to inspect and evaluate. Commercial and production use requires a license.
A commercial license grants production use and includes curated live pattern feeds, calibration against labeled attack corpora, and support from the person who built it.
Because NV is self-hosted, the license adds capability and support on your own infrastructure — never a requirement to send your data anywhere.
💼 Commercial Licensing
Early access. NV is pre-1.0, and I'm onboarding a small number of founding users on real investigations. Founding licenses are one-time and perpetual — pay once, keep it, no renewals — at introductory pricing in exchange for feedback. Each includes 12 months of feed and pattern updates, with optional renewal after that to keep the intel current. Request a license and I'll send current terms and set you up with the live demo.
| Tier | Founding License | Which one am I? |
|---|---|---|
| Responder | from £490 · one-time | I run investigations solo or as a consultant |
| IR Team | from £2,400 · one-time | My security team uses it on live cases |
| Enterprise | Let's talk | We're an MSSP or large org deploying it broadly |
< Install & Use My Open Tools />
Most of what I build is free and open. Install it, use it in your own work, or support it if it helps you.