Company

A research-led company building trustworthy, accountable AI for high-stakes industries.

SparkSoft builds the assurance layer for applied AI from London — selected by Innovate UK and working with UK institutions across manufacturing and bioscience.

Founder

Engineer, researcher, founder.

SparkSoft was founded in London by Dr Peng Yue, who has led it for six years — pairing more than twenty years of building software and enterprise systems with a doctorate completed this year. The conviction behind the company comes from both halves of that career: a strong benchmark result is useful, but it does not make a product deployable. Industrial AI has to be grounded in representative data, the real process it joins, clear acceptance criteria and feedback from the people who carry responsibility for the outcome.

Over the last four years SparkSoft has won six UKRI and Innovate UK projects, worth around £370k combined — from climate-smart pest management to AI-driven identification of crop pests, built with the University of Sheffield and The Alan Turing Institute — before the £770k Sovereign AI award behind this platform. The through-line in every project is the same: AI that can show its evidence.

Dr Peng Yue on LinkedIn →

The programme

Deployable and Scalable Symbolic Capsule Neural Networks for Auditable Zero-Shot Object Detection

Selected by Innovate UK under Sovereign AI Pipeline Phase 2 — one of 17 companies across the UK, backed by a grant of around £770k. The programme funds our core research: detection designed to flag defect types it was never trained on, with every call auditable. With AMRC we are now identifying manufacturing use cases and design partners for a demonstrator that proves technical performance and business value together.

Funded by

Innovate UK UK Research and Innovation

Grant record available on request for due diligence.

How we work

Design partners, not an abstract market.

Start small, prove it

A partnership starts with one workflow, agreed data, explicit assurance questions and success criteria — and earns its way to scale.

Measured, not promised

Every pilot begins with operational and financial baselines, so the outcome is a measured time, cost or quality benefit — not a demo that flattered the lab.

Claims we can stand behind

We don't market assurance claims that haven't been validated through evidence, testing or external review. The website you're reading follows the same rule.

PARTNERS & COLLABORATORS
AMRC The Alan Turing Institute University of Sheffield Fera Science Harper Adams University ADAS

Research

The citable record.

The dissemination record of our Innovate UK project — the questions we are working on and what we will publish as the work matures.

Read the research record →

Visit

6th Floor, First Central 200
2 Lakeside Drive, Park Royal
London NW10 7FQ

Short walk from Hanger Lane (Central) and Park Royal (Piccadilly).