AI evaluation startup Trismik raises £2.2m led by Twinpath Ventures
At a glance
This article covers the pre-seed round on 24 September 2025 for Trismik, a Cambridge-based AI startup applying psychometrics and adaptive testing to evaluate large models, founded by Rebekka Mikkola and Nigel Collier. The company raised £2.2m in a round led by Twinpath Ventures with participation from Cambridge Enterprise Ventures.
What does Trismik do?
Trismik offers an AI evaluation platform that uses adaptive testing to map model capabilities. It is used to generate granular assessments for model selection, compliance reporting and experiment analysis.
Why is Trismik useful?
Problem
AI teams cannot reliably measure large models because popular benchmarks are saturated and uninformative. This prevents precise model selection, domain-specific testing, and clear communication of capabilities to stakeholders.
Solution
Trismik explains that it applies psychometrics, Item Response Theory and adaptive testing to evaluate models more precisely. Its adaptive tests map model ability efficiently, cutting questions and cost while giving nuanced, interpretable results.
How much did Trismik raise?
Trismik raised £2.2m in a pre-seed round led by Twinpath Ventures, alongside Cambridge Enterprise Ventures. This makes it the 39th largest funding round in September 2025 (47 recorded). The deal is also 303rd among UK investments this year (454 recorded) in the Startupmag database, as of 24 September 2025.
For details on how Startupmag compiles its rankings, view our Methodology.
Who invested in Trismik?
Key investors in the round included the following:
- Twinpath Ventures: A venture capital firm that invests in early-stage technology companies, including AI and deep-tech startups.
- Cambridge Enterprise Ventures: The University of Cambridge's investment arm that backs spinouts and deep-tech ventures emerging from the university.
- Parkwalk Advisors: An investment firm that specialises in supporting UK university spinouts and technology companies.
- Fund F: A venture capital investor that backs early-stage tech companies, often focusing on software and data-driven businesses.
- Vento Ventures: A venture capital firm investing in early-stage technology startups across Europe.
- Ventures Together: Angel investors who participated via the Ventures Together syndicate, with interests in technology and AI.
In the funding announcement, John Spindler from Twinpath Ventures said:
The AI evaluation market is at an inflection point. Every AI team we speak with is drowning in evaluation overhead, it has become the hidden bottleneck preventing teams from shipping faster and with confidence.
The investor said Trismik's approach applies proven scientific methods from another domain to address evaluation challenges and that substantially reducing evaluation time while improving measurement precision can change what is possible in AI development cycles.
If you're researching potential backers in this space:
- Explore our list of AI venture capital firms in the UK
- Browse our directory of AI angel investors in the UK
Who founded Trismik?
The founders of Trismik are Nigel Collier and Rebekka Mikkola.
In the funding announcement, Nigel Collier, Chief Scientific Officer of Trismik said:
“If we want to trust AI, our methods have to be as rigorous as our ideas. Benchmark saturation is creating problems in every domain, from general knowledge, to reasoning, math, and coding.
The company continued that scientists, researchers and technical teams face mounting pressure as evaluation becomes increasingly important for tying AI to trust, and that there is a need for an evaluation framework that scales and can support this.
Where is Trismik based?
Trismik is based in Cambridge, UK.
What sector is Trismik in?
Trismik operates in the AI sector. It develops systems that can perform tasks requiring human-like intelligence. In plain terms, AI helps computers understand language, solve problems, and learn from data.
Key trends and challenges in AI:
Benchmark saturation and capability measurement
Popular tests like MMLU and GSM8K show many models scoring above 90 per cent, reducing useful differentiation.
Domain-specific evaluation and compliance pressures
Generic benchmarks often miss sector needs, for example in healthcare or finance, complicating safety checks and regulation.
Rise of adaptive, psychometrics-based testing
Researchers are applying psychometrics and adaptive testing, sometimes needing only around ten per cent of questions to assess models.
For a deeper look at innovation in this space, see the latest funded AI startups in the UK
They've invested in Trismik
Get to know these AI investors
Click here for a full list of 7,233+ startup investors in the UK