Literal Labs

Literal Labs is a UK artificial intelligence company that develops Logic-Based Networks (LBNs), a model architecture that replaces the floating-point multiplication used in neural networks with propositional logic while retaining deep learning training. The approach derives from Tsetlin machine research and related work on data binarisation, model compression and parallel hardware implementations carried out at Newcastle University and referenced alongside the University of Agder's Centre for AI Research.

Its product is ModelMill, a training platform that ingests datasets (CSV, JSON or ZIP up to 5GB) via browser or API, handles pre-processing, normalisation and annotation, lets users specify target deployment hardware and optimisation priorities (energy, speed or memory), trains and benchmarks hundreds of LBN candidates in parallel, and packages a selected model as a C SDK with an inference engine, build configuration and documentation. The resulting models are described as running on any 32-bit processor — ARM, RISC-V, x86, ESP and PowerPC — without GPUs, new silicon or a cloud connection, with average model size under 40kB and accuracy within ±2% of larger GPU-dependent algorithms. The company states LBN behaviour is deterministic (identical input yields identical output) and traceable for audit, which it positions as relevant to regulated industries.

Published use cases and benchmarks span the water industry (battery-powered sewer sensors running hydro-informatics forecasts), automotive edge AI on PowerPC, supply chain inventory forecasting, predictive maintenance and semiconductors. The company reports being spun out of Newcastle University, is registered at Companies House under number 14746541 and holds D-U-N-S number 230470826.

Founding story

Literal Labs was created out of Newcastle University research by Professor Alex Yakovlev, head of the Microsystems Group, and Professor Rishad Shafik, EEE Research Director, together with venture builder Cambridge Future Tech. The founders' work covered advanced compression methods, data representation and highly modular, parallel implementations of the Tsetlin machine, a logic-driven alternative to neural networks inspired by the work of mathematician Mikhail Tsetlin. Noel Hurley, previously more than two decades at Arm where he led the CPU division, was appointed CEO in 2023. The company's own about page dates the spinout to 2023, while Newcastle University's press office describes it as spun out in 2024.

Business model

Literal Labs sells access to ModelMill, a cloud-based AI model training platform, with trained models delivered as an embeddable pure-C SDK containing the inference engine, build configuration and documentation for embedded or server deployment. Early access to the platform is offered through a request form on the company website.

Traction

Literal Labs reports LBNs in production across automotive, utilities, supply chain and semiconductor applications, with published deployments including battery-powered sewer sensors forecasting every five seconds on coin-cell-equivalent power for over ten years (replacing an LSTM that would have required roughly £15,000 of energy capex per sensor), predictive maintenance models cutting unplanned downtime by up to 50% and maintenance costs by up to 40%, and inventory forecasting reduced from four hours to three minutes with 2× WMAPE accuracy across thousands of SKUs. Headcount doubled from six to twelve during 2024.

Latest developments

Company updates list Jim Darragh's appointment as Non-Executive Chairman (March 2026), an AI Research & Innovation Award win (March 2026), selection as one to watch in Barclays' AI 100 (May 2026), and a scheduled Edge AI London appearance (June 2026). The leadership team now also includes CFO Mike Park and Head of Product Daniel P Dykes, with a board and advisory group including Professor Ole-Christoffer Granmo as Technical Steering Committee chair, Jem Davies as Non-Executive Director, Kelvin Harrison as board observer, and investor representatives from Mercuri, SVV and Northern Gritstone.