What We've Learned

Our customers are innovators, pushing the boundaries of science and engineering to create and optimize materials, chemicals, products, and processes. They generate intellectual property, which is an intangible, but no less important, asset on the balance sheet. It is important that this asset is protected.


This starts with documentation. Not the most thrilling topic, but proof is needed on who invented what, when. Even in territories that grant patents based on filing date rather than invention date, it is important to establish that IP was created by employees under contract. Historically, this has not been as easy to establish as you would think. Researchers’ coffee-stained notebooks and disperse spreadsheets needed to be hunted down. Now with systems like the Citrine Platform companies can easily look up who did what at what time digitally.


Questions have been circulating about the ability to patent work created with AI help. As the new guidance on inventorship for AI-assisted inventions from the United States Patent Office (USPTO) says, “the inventorship analysis should focus on human contributions, as patents function to incentivize and reward human ingenuity.” AI can’t hold a patent. To get a patent a “natural person” must have “provided a significant contribution to the invention”.

How does that work in the Citrine Platform? Since its foundation in 2013 Citrine has been focused on a world of small data. The companies that Citrine works with have data sets ranging from 20 – 1000 data points, as each data point can cost $1000’s to acquire. “Big Data AI” doesn’t work here. It has therefore been very important that the Citrine Platform leverages the expertise of customer teams to structure AI models and focus it on searching design spaces that are practically feasible and likely to be successful.

Although our customers save up to 80% R&D time to achieve their target properties, this is because fewer experiments are needed, not because researchers are not involved. Researchers using the Citrine platform:

  1. Decide which data is important to the project
  2. Use a graphical interface to tweak an automatic first-pass AI model; ticking boxes to indicate which inputs are related to which output properties; inputting scientific equations that can give the model insights impossible to glean from the data…
  3. Choose which ingredients and how much of them are feasible from a business and practical perspective
  4. Set targets and constraints on final properties
  5. Review and decide which of the suggested candidate experiments they should run

It is therefore clear that the researcher still contributes significantly to any patentable results.


Citrine is very conscious of its duty to protect the IP of its customers. We’ve been ISO 27001-certified since 2018, and have continued to invest in our information security beyond ISO requirements. As we start to enable our customers to use private large language models (LLMs – like ChatGPT but private) it was very important to us that customer data remain secure. You can read in more detail how security is at the heart of everything we do here.

Business Model

Our customers create IP using the CItrine Platform. Just like they use SAP, or Excel, this boosts productivity but does not give Citrine Informatics rights to the IP produced. While some consultants using AI might consider a different business model, Citrine is clear, our expertise is in software and AI. We are proud of our patents in this area.

Intellectual property is generated, documented, and subsequently patented by researchers using the Citrine Platform. Citrine is proud of the work our platform supports.