About Liqa Labs
Liqa Labs works on two sides of the same problem: applied AI engagements with enterprises, and research data for frontier labs. We help companies put today's models to work, and build the environments that train the next ones.
Applied: move real workflows onto AI
Most AI projects automate a process as it stands. The gains come from fixing the process first. We start inside the workflow: how the work is specified, revised and judged. We rebuild it with the team, then decide where a model belongs and where it doesn't.
Engagements are short and run by people who have done the work. The same evaluations we build for labs tell you which models and configurations meet your standard.
Research: RL environments for knowledge work
We build human expertise and best practice into how automations are trained and judged. Our tasks and graders tell apart what holds everywhere from what is specific to one organisation, and extract the second part from that organisation's own prior work.
This work has been developed with more than 100 companies and an expert network of thousands of professionals.
Benchmarks, datasets and field notes
Doppel Benchmark tests whether frontier models can reproduce a company's visual intelligence system: its published brand guidelines and the conventions its teams actually work to.
Doppel Dataset is a proprietary collection of real-world corporate presentations and prompts, created by human experts with enterprise backgrounds and available under commercial licence.
System One models vs. LLMs compares specialised decision models against lightweight LLMs on picking the right visual for a slide.
Company
Liqa Labs is built by a team of researchers, engineers and operators, and was previously known as Tesserae. Liqa Labs is a trading name of TryTesserae Ltd, registered in England and Wales, company number 16256255. Registered office: 128 City Road, London EC1V 2NX, United Kingdom.
Our security and compliance posture (SOC 2, ISO 27001 and GDPR) is documented in our trust center. How we handle personal data is set out in our privacy policy. To get in touch, see contact.