iKVA embarks on £2.5m round to enhance AI-driven vector mapping
iKVA has embarked on a pre-Series A finance round for £2.5million to accelerate the development of its AI-driven knowledge management software solutions.
iKVA’s “genesis” in 2018 involved three academics from the university’s Department of Computer Science and Technology, and the Alan Turing Institute – Dr Liang Wang, Professor Richard Mortier and Professor Jon Crowcroft – plus Jon, whose backstory includes stints at Cambridge Assessment and Local Heroes, a British Gas digital innovation.
The latest funding round – “a pre-Series A round is also known as a seed-plus” – follows on from £1.5m seed investment raised exactly a year ago. iKVA has now raised a total of £1.675m – a first round at £175,000 and the second round of £1.5m – from investors including Cambridge Enterprise and BTF.
iKVA’s cloud-based technology works by pulling together unstructured data from multiple internal and external sources – including rich documents, images, video and chat – to provide accurate knowledge to the user in real time as part of their regular workflow.
The early-stage company’s platform uses vector analysis, which CEO Jon Horden describes as context-driven rather than keyword-driven.
“We use vector mapping technology,” he explains in a Zoom call.
“The AI is mainly in the neural networks and proprietory software. It’s looking for patterns and context, so we understand the context of a document – not just a Word document, it could be a tweet, speech, or post – and we convert the document to a bunch of vectors. These are 300-dimensional vectors in mathematical space. If you try and convert the vector to 3D, a document becomes a ball of points in space, and you’re looking for similarities and relevance.
“Fundamentally, we’re doing mapping, we use mathematical techniques to look at this data, and a lot of information surfaces which you wouldn’t get if you rely on keywords, which is what Google uses.
“Our clients have very, very complex data – for instance with scientists or product engineers – and keyword searching isn’t enough. We understand the entirety of the document, plus the software is completely language-agnostic, it all goes into mathematical space so you can do a query on an Italian document and get the result back in English.
“In the old world you’d have to do keyword searches in different languages – we don’t do translation, we use neural networks trained across multiple language sets.”
Jon says he “can’t take the credit” even though he is one of four founders of the company which is a finalist in two categories of the Cambridge Independent’s Science & Technology Awards 2022 – AI Company of the Year and Start-Up of the Year.
“We’re fairly early on at this point,” says Jon of this latest investment round. “Usually you’d hope to get it done in three months. It will almost all go on product development – 70 per cent on research and 30 per cent on sales and marketing.”
The potential is vast.
“We can track the metaverse to before it even had a name,” says Jon. “There was no such term five years ago, but we understand concepts and context without using keywords, so we can say to a client ‘there’s a cluster of things going on here, and here’s some terms being used in the cluster’.”
Clients including Mott MacDonald have been suitably impressed.
“We hope the clients will be around forever, though typically they sign up for a three-year contract.
“We’re an early-stage company, and lots of big companies are coming on board.”