Intelligence, Accelerated

Recurvia is building the intelligent systems and tooling that will unlock the next generation of self-optimising AI.

Illustration of a Gaussian Map Surface

Built by researchers and engineers from

Apple Logo
University of Cambridge Logo
Harvard University Logo
University of Oxford Logo
GResearch Logo
Instadeep Logo
NeptuneAI Logo
Bloomberg Logo
Our DNA

From dynamic training control to AI-native experimentation, our systems are designed around continuous optimisation rather than static workflows. This research DNA runs through everything we build.

THE PROBLEM

AI is entering an optimisation era.

As models continue to scale, compute requirements are becoming one of the defining constraints in AI.

Isometric illustration of a microprocessor unit

Building AI that can keep up

Training and deploying advanced AI systems increasingly depends on iteration velocity and compute efficiency - at all levels of the stack

3D Illustration of a convolution layer diagram

The next AI won't wait to be told what to do.

At the same time, global demand for AI compute continues to accelerate, creating growing pressure across GPU supply chains, memory systems, and large-scale infrastructure.

The next generation of AI systems will not simply generate outputs. They will continuously analyse, optimise, and adapt themselves.

Core Technologies

From dynamic training control to AI-native experimentation, our systems are designed around continuous optimisation rather than static workflows. This research DNA runs through everything we build.

PARAMORPH

A foundation model for neural network optimisation.

Paramorph dynamically adapts optimisation strategies during training — continuously adjusting hyperparameters and training behaviour at runtime to improve convergence efficiency, scalability, and system performance. 

Built using adaptive control and multi-agent learning techniques, Paramorph approaches optimisation as an intelligent, continuously evolving process rather than a static configuration problem.

LEARN MORE
Isolated Product Mockup
Clock Icon

Dynamic optimisation at runtime

Convergence Icon

Faster training convergence

Parameters Icon

Reduced hyperparameter sweep overhead

Bot Icon

Multi-agent coordination

Layers Icon

Layer-wise adaptive control

Metrana

AI-native observability for modern machine learning workloads.

Originally developed internally to support Paramorph's optimisation workloads, Metrana evolved into a next-generation observability and experimentation platform for increasingly complex AI systems.

Metrana combines large-scale metric logging, intelligent analysis, agentic experimentation, and adaptive infrastructure workflows in a platform designed for modern AI research and development.

LEARN MORE
Isolated Product Mockup
Stopwatch Icon

Metric logging at massive scale

Search Icon

RL-specific observability interfaces

Sparkle Icon

AI-assisted analysis

Repeat Icon

Automated experimentation and iteration

THE CHALLENGES

Optimisation is becoming a defining AI capability.

As frontier AI systems scale toward ever larger training runs and increasingly complex infrastructure, even small inefficiencies can translate into millions of pounds lost.

Wallet Icon

Millions in compute cost

Clock Icon

Extended training timelines

Meter Icon

Reduced iteration speed

Server Icon

Infrastructure underutilisation

BACKED BY LEADING DEEP-TECH INVESTORS AND ACCELERATORS

Sure Valley Ventures Logo
Amazon Web Services Logo
Amadeus Logo

The future of AI will be self-optimising.
We're building the systems to power it.