Matterhorn Studio: Your first step towards AI-driven sustainable materials development (with a focus on scale-up of bioengineering)

通过在机器学习(ML)的帮助下智能地安排实验，Matterhorn加速了资源节约型材料设计。诞生于伦敦大学学院的人工智能中心，我们为材料公司提供了在机器学习驱动的实验设计(DOE)方面取得世界级学术进步的机会。我们提供免费的软件，与用户友好的实验管理数据平台无缝集成(“马特霍恩工作室”，http://matterhorn.studio).Historically，开发资源高效的材料是一个由理论和直觉驱动的过程。不可避免的是，由于材料的日益复杂，这种理论分析和基于直觉的实验的回报正在减少，从而增加了开发成本，但成功的次数更少。能源部方法的最新进展在实验室革命中发挥了核心作用，例如，制药行业可以实施实验的“闭环系统”，达到马修·里夫的数字成熟度框架的“4级”。一个完全自动化的“4级”实验室使科学家们可以自由地从事其他更复杂的任务。最近的报告声称，材料开发速度加快了10到100倍，成本降低了10到100倍(acceleration.utoronto.ca/)。不幸的是，大多数实验室负担不起实现“闭环”实验所需的投资。马特霍恩使这些实验室能够升级到1级和2级。我们观察到，对于希望将实验室升级到更高水平的早期采用者来说，聘请一名数据科学家通常是第一步。这些数据科学家是马特霍恩的主要受益者，因为它解决了他们决定使用哪种算法以及如何安全管理数据的问题。展望未来，我们希望与更广泛的材料社区分享马特霍恩在资源高效材料方面的努力。这笔赠款将通过帮助我们开发我们的平台，帮助材料科学家在机器学习方面迈出他们的第一步，从而帮助实现这一目标。马特霍恩将为生物工程或固态化学等广泛的材料领域提供专用模型。在易于使用的平台和可访问的教程的帮助下，我们希望激励和支持下一代材料科学家发展他们在数据驱动的材料发现方面的技能，并推动英国和全球材料生态系统的整体进步，同时提供一个专门的平台来照顾他们的数据管理、实验时间表和团队协作。

Matterhorn accelerates resource efficient materials design, by intelligently scheduling experiments with the help of Machine Learning (ML). Born out of UCL's AI Centre, we give materials companies access to world-class academic advances in Machine Learning driven Design of Experiments (DOE). We provide a freely available software that seamlessly integrates with a user-friendly data platform for managing experiments ("Matterhorn STUDIO", http://matterhorn.studio).Historically, developing resource efficient materials is a process driven by theory and intuition. Inevitably, due to the ever increasing complexity of the materials, returns from such theoretical analysis and intuition-based experimentation are diminishing, therefore increasing development costs with fewer successes.Recent advances in DOE methods have played a central role in revolutionising laboratories, for example, pharmaceutical industries can afford to implement a "closed loop" of experimentation, reaching "Level 4" of Matthew Reeve's Digital Maturity Framework. A fully automated "Level 4" laboratory frees the scientists to work on other more complex tasks. Recent reports claim 10 to 100 times faster materials development with a 10 to 100 times reduction of costs (acceleration.utoronto.ca/).Unfortunately, most labs cannot afford the investments required to achieve "closed loop" experimentation. Matterhorn enables these labs to upgrade to Level 1 and 2 instead. We have observed that hiring a single data scientist is often the first step for early adopters, that want to upgrade their labs to higher levels. These data scientists are the main beneficiary of Matterhorn, since it solves their problem of deciding which algorithm to use and how to securely manage the data.Moving forward, we would like to share Matterhorn with the wider materials community in their efforts towards resource efficient materials. This grant will help make that possible by helping us develop our platform where materials scientist can make their first steps in machine learning. Matterhorn will provide dedicated models for a wide set of material fields such as bioengineering or solid-state chemistry. With the help of an easy to use platform and accessible tutorials, we hope to inspire and support the next generation of material scientist to develop their skills in data-driven materials discovery and advance progress in the UK and global materials ecosystem as a whole, while providing a dedicated platform to take care of their data-management, experimentation schedule and team collaboration.