About Trismik

Building the Future of AI Model Selection & Optimisation

Trismik helps teams choose and optimise the right language model with evidence they can trust.

We’re building a model decision platform for AI teams making real trade-offs between quality, cost, and latency. Instead of relying on leaderboards or ad hoc scripts, Trismik makes model choices repeatable, measurable, and easy to justify - internally and to clients.

QuickCompare is your model decision workspace

In minutes, teams can compare LLMs on the prompts and data that matter to them, surface failure modes, and make confident decisions about which model to ship.

The Journey We Enable

Together, these capabilities help teams move from initial decisions to ongoing optimisation

Step 1
"Which model should we use?"

Step 2
"How do we optimise performance and manage trade-offs over time?"

Meet the Team

The people behind Trismik

Rebekka Mikkola

CEO & Co-Founder

Repeat founder and former Salesforce executive.

Nigel Collier

CSO & Co-Founder

Cambridge NLP Professor and 30+ years of experience in AI.

Marco Basaldella

CTO & Co-Founder

PhD in AI and former Amazon applied scientist.

Our Story

Trismik began at the University of Cambridge, where Professor Nigel Collier saw first-hand how slow and operationally painful Large Language Model evaluation had become.

Inspired by Computerized Adaptive Testing from psychometrics, he asked a simple question: could models be evaluated with the same efficiency and statistical rigour as humans? That idea became the foundation of Trismik's adaptive evaluation approach.

In 2023, Nigel partnered with Rebekka Mikkola, a repeat founder and former Salesforce enterprise sales executive, to build Trismik's first MVP with early backing from Cambridge Enterprise and an early design partner in the UK.

In 2025, Marco Basaldella, Nigel's former postdoc and an ex-Amazon scientist, joined as CTO.

Our Vision

AI models will keep evolving, and teams will increasingly rely on multiple models across workflows. Model choice can't be a one-off decision - it needs to become a continuous, evidence-based practice.

Our goal is to become the trusted platform for AI model decision and optimisation: helping teams compare models, manage trade-offs, prevent regressions, and continuously improve performance as their systems and requirements evolve.