London-based
Hypercritical has raised £2 million in pre-seed funding to accelerate
development of its foundation model and expand its engineering team. The round
was led by Join Capital, with participation from Octopus Ventures, Tiny
Supercomputer Investment Company (tiny.vc), and Plug and Play.
Hypercritical
is a deeptech company that develops machine learning models to generate fully
correct control software for safety- and mission-critical systems. Using a
novel, logic-driven architecture that eliminates hallucinations and errors, it
enables engineers in sectors such as automotive, aerospace, defence, and
robotics to design and deploy control systems more quickly and
cost-effectively, with mathematically guaranteed reliability.
Instead of
writing code directly, engineers define the tests a system must satisfy, and
Hypercritical’s AI automatically generates algorithms that meet all of them,
enabling fully automated control development for highly demanding physical
systems.
Its Copilots deliver immediately usable, domain-specific output, while
its Autopilot produces unsupervised software that passes 100% of tests. This
results in software generation that is significantly faster, more
cost-efficient, and mathematically precise, essential in industries where errors
are unacceptable.
Hypercritical
aims to make its technology the benchmark for generating control software in
safety- and mission-critical systems, and ultimately envisions its methods
being incorporated into ISO standards to help modernise global software
certification and compliance.
The
company’s flagship product, Hyperpilot, is already in use by engineering teams
to automate the development of systems that rely on control software. In
parallel, key components of its technology stack, including domain-specific
“copilots” such as a QA engineer and a systems engineer, have been deployed
with customers, demonstrating its applicability in real-world, safety-critical
environments.
Following the raise, Hypercritical plans to double its
team, with funds primarily allocated to hiring and to cloud compute for
training its proprietary model.
Image: Freepik