ElverElver
HiringForward Deployed Engineer

Elver · Applied AI · Bristol, UK

AI research products
for mission driven organisations.

Consolidate and compound your research with AI products built to fit established workflows.

What is custom AI?

A product built for you

Uses AI intelligently, everywhere

Improves automatically with use

Never locked to one model or vendor

Systems, not seats

Elver is a UK-based AI consultancy. We build client-owned products that we can host and manage ourselves, or integrate fully into client infrastructure.

AI is at the core of our approach, with a clear separation between probabilistic and deterministic components.

Flywheel effect

human input + AI orchestration → flywheel effect → institutional IP

Fig. 1 — the flywheel between your people and your AI.

How we build

Sargasso

Our own agentic management system. It's an evaluation harness, it orchestrates agents, it's also a delivery pipeline and a client portal. Sargasso is not a product we sell, it's how our small team can ship high-quality software so efficiently. It's also how we can ensure everything we build is safe and secure.

sargasso · factory floor

replay

Fig. 2 — a Sargasso job.

Live Demo

GSS Explorer

As most of the products we build are not public facing, we produced this example. Powered by the General Social Survey, an US attitudinal survey that's been running since 1973, it allows you to describe a story and an agentic system will isolate relevant data. Our storytelling engine extracts themes from your narrative and builds charts that support, or challenge, your case. The user requires no knowledge of the underlying survey data. Built using Sargasso in two days.

Fig. 3 — GSS Explorer: describe a narrative, get fifty years of survey data to support or challenge it.

Try it yourself →

Co-development

All our AI systems are co-developed with our customers. AI is a fast-moving field with baffling jargon and we upskill the teams we work with. We are also well placed to help clients work through the ethical implications of AI adoption.

The vocabulary, demystified together

01 — Parametric knowledge

What the model already knows from its training, baked into the model itself. Nothing is looked up. It answers the way you'd recall your own phone number.

Fig. 4 — Demystifying the vocabulary of AI. Hover or tap any term to hold it and read what it means.

Company

Fifteen years of building together.

Elver was founded by Nick Fellingham and Andy Littledale, who have worked together since 2011 across three startups. Their first was an analytics business acquired by Twitter, the second a VC backed deep-tech video firm. Elver was started in December 2025 when it became clear that AI models would fundamentally change how software was developed.

Nick Fellingham

Nick Fellingham

Co-founder · Technology

AI specialist with deep expertise in large language models and generative systems. Previously CEO and co-founder of Condense, the VC-backed volumetric-video company behind immersive broadcasts with the BBC; before that, engineering at Twitter and a decade of applied ML, starting with the systems at SecondSync that mapped global conversation to live television in real time.

X @Nick_Fellingham →LinkedIn →

Andy Littledale

Andy Littledale

Co-founder · Commercial

30 years working in tech. Co-founded SecondSync, the social-TV analytics company acquired by Twitter, then led audience-analytics products at Twitter. Founded Poli, a micro-polling startup rethinking how public opinion is measured. Honorary Professor at the University of Bristol, where he lectures on innovation and entrepreneurship. Philosophy degree. Advisory board member of Practice-Oriented AI CDT.

LinkedIn →

Now hiring · Bristol · £120k

Forward Deployed Engineer

Half with customers, half building. Own the work from the first conversation to production, with the founders sitting next to you.

Read the role →

Contact us

Get in touch.

If you'd like to talk about a project or just ask us a question, please get in touch using the email or WhatsApp details below.