Run the best possible experiment. Every time.

Danu integrates your team’s expertise with cutting-edge AI, ML and information theory, so every next experiment is the most valuable one you can run.

  • —

    Learn more from every experiment. Information theory instead of trial-and-error.

  • —

    48 hours to first insights. We think nobody does this faster.

  • —

    Your data and your process. Starting from your problem, not a template.

Danu started with one ambition: to bring breakthrough information theory and AI to biotech. We want to put these to work on the hardest problems with the biggest potential impact, from disease and health to sustainable materials.

We build our technology together with the scientists who use it. Through our novel, award-winning methods, we have enabled our partners to dramatically reduce the number of runs needed to reach a successful result, and to drastically speed up their R&D progress.

Our mission is simple: accelerate biotech discovery, scale-up, and commercialisation, so the breakthroughs that matter reach all of us sooner.

Working with Danu

An illustrative project with Danu: from setting goals to optimising experiments to transferring knowledge into the next campaign.

Click any step to jump to it, or press play.

0 MONTHS: YIELD 0.8 G/L6 MONTHS: YIELD 5.2 G/L12 MONTHS: YIELD 7.9 G/LYOUR SCIENTISTStacit knowledgeHISTORICAL DATA14 past runsLITERATURE6 papersSIMULATED · BEFORE ANY WET-LAB EXPERIMENTCONVENTIONALMODELNEXT CAMPAIGNwith a head start
GOAL: YIELD 7.0 G/L
PROPOSES
WET-LAB RESULTS
YOU KEEPA better processAn integrated datasetProcess insights to kick-start new campaigns

Case studies in numbers

Productivity climbing 66 percent from the standard medium across four optimisation roundsProductivity climbing 66 percent from the standard medium across four optimisation rounds

+66% productivity through medium optimisation

We optimised thirteen medium components simultaneously to boost productivity for a high-value chemical, improving it by two thirds over four rounds.

Many scattered growth-curve experiments feeding into a single model of final cell density against one medium component, with its uncertainty bandMany scattered growth-curve experiments feeding into a single model of final cell density against one medium component, with its uncertainty band

A predictive model without a single new experiment

We combined dozens of historical runs with hundreds of components that were never designed to be compared. We unified them into one dataset, trained a single model, and used it to kick-start the next campaign.

Final regret against random search: Danu 0.45, LLM agent 0.67, DoE and RSM 0.89, random search 1.00Final regret against random search: Danu 0.45, LLM agent 0.67, DoE and RSM 0.89, random search 1.00

Twice as close to optimum as DoE

In extensive benchmarks, Danu’s decision engine finished twice as close to the optimum as industry-standard Design of Experiments at the same budget, outperforming classic methods and frontier LLM agents.

Questions we get asked

Software, a service, or both?

Both, we provide cutting edge software with a whole lot of love and care. This includes understanding and utilising what you already have, getting you set up AI-ready for any future projects, and providing bespoke insights on noise, variance, and failure drivers.

What’s a Data Jam?

We want to prove, at no cost to you, what we can do for one specific R&D problem you have in just 48 hours. One call with your scientists on the Monday. Tuesday and Wednesday you gather whatever data exists and sort the NDA. We jam Thursday and Friday. Results the following Monday.

How is it priced?

The Data Jam is free for you, no commitment, no cost. Beyond that, we price per project. We don’t do per-seat licenses, as we believe anyone who should get access can.

Who owns the data, insights, and process improvements?

You own all data sets and every insight and process improvement that comes out of the work.

Why not use DoE software or a modelling team?

DoE hands you an off-the-shelf design. A modelling group hands you a bespoke model in months. We produce a bespoke model in days, and twice as accurate as standard DoE on the same budget. In addition, make sure your experimental setup and data collection is AI-ready, no matter who you choose to work with in the future.

More questions and answers