Fuzzy Labs

What they do

Fuzzy Labs builds and productionises AI models for enterprise clients, with open-source MLOps as the throughline. The work spans the full pipeline: improving how data scientists collaborate, speeding up model deployment, and implementing model monitoring. They describe their approach as "practical AI, properly engineered" — a fair summary of a team that includes multiple PhDs in astrophysics, particle physics, experimental particle physics and machine learning, alongside engineers with backgrounds in defence, logistics, manufacturing and the public sector.

Case studies on their site point to work including building an AI assistant to halve case file preparation time, onboarding new polytunnels in days rather than weeks, and uncovering critical flaws in AI security tooling. The public sector is a named vertical: Jamie Smith, their Public Sector Lead for Policing, spent 15 years at South Yorkshire Police before joining, and the team explicitly references responsible AI standards as a constraint on that work.

The consultancy publishes a cookbook — Cooking with MLOps — described as practical templates for building AI systems with repeatable results. Blog output covers agentic AI, LLM cost analysis, and responsible deployment, with named authors including CTO Matt Squire and engineers across the team. They also maintain open-source projects, though specific repository names are not listed on the public pages.

Contact and billing go through their Manchester office at Heron House, 1 Lincoln Square, inside the GM Digital Security Hub (DiSH). The phone number is a Manchester 0161 prefix. There is no Leeds office listed.

What sets them apart

The core argument Fuzzy Labs makes is straightforward: proprietary MLOps platforms lock you in, open-source tools do not. They make this case explicitly on their about page, noting that bespoke open-source solutions are faster, cheaper and more flexible than vendor platforms — and that clients own the result. It is a position that requires some confidence to hold in a market where the major cloud providers are pushing their own managed ML services hard.

The team is notably research-heavy for a consultancy of this size. Three MLOps Tech Leads hold PhDs (astrophysics, particle physics, and one further physics background), and the Lead AI Research Scientist, Dr Danny Wood, spent eight years at the University of Manchester working on recurrent neural networks, ensemble learning, uncertainty quantification and explainability. That depth shows up in the blog, which engages seriously with questions like agent reliability in production and the cost of autonomous systems.

Robbie Vigers, CCO, brings five years of AI-focused commercial experience from Peak AI before joining. The operations side is handled by Tim Carr, who has over 20 years in operations across translation and digital marketing. Ian Brookes, investor and advisor, has 30 years as a startup founder and investor and helps with commercial and financial frameworks.

The team also runs internal hackathons — a September 2025 blog post documents building TROLL, an AI-assisted ticket-writing tool, as an internal experiment. That kind of self-directed tinkering, alongside the open-source contributions, suggests a team that uses the tools they sell rather than just talking about them.

What they offer

• Open-source MLOps consultancy — bespoke model deployment and monitoring pipelines built with open-source tooling, positioned as an alternative to proprietary vendor platforms.
• Data science team embedding — Fuzzy Labs works alongside existing in-house data science teams rather than replacing them, handling productionisation and MLOps infrastructure.
• Public sector AI deployments — specialist work in policing and criminal justice, led by Jamie Smith, with a focus on responsible AI standards and measurable productivity gains.
• MLOps cookbook (Cooking with MLOps) — a published guide of practical templates and repeatable recipes for building AI systems, available from their site.
• Open-source projects and documentation — maintained resources available via their Labs section, though specific project names are not listed on the public pages.