Ready to learn more?
Request a demo to learn how your team can accelerate battery R&D
Blog
Conducting Smarter Battery Science with Data Management and AI
No items found.
Read a recap of our recent presentation at The Battery Show Europe and watch our interview with Charged EV Magazine.
We recently joined over 1,000 exhibitors and 20,000 attendees at The Battery Show Europe in Stuttgart, Germany. As we connected with the brightest minds in the battery industry, a shared theme emerged: developing battery materials is inherently complex and entails large quantities of data. Moritz Haus, PhD, Solution Engineer at Albert Invent, spoke with Charged Electric Vehicle Magazine about how AI trained like a chemist can harness complex data to accelerate battery R&D:
Transforming battery complexity into battery clarity
Moritz took the stage and dove deeper into this topic through his presentation “Data Management and AI for Next-Generation Battery Materials Development.” He opened with an overview of a typical battery research workflow, revealing the reality that battery development is not just multi-step, but also multivariate.
“With so many factors at play, it becomes incredibly challenging to trace performance back to your formulation or process variables,” said Moritz. “Especially when that data is recorded across dozens of spreadsheets, in different formats, by different people.”
Moritz admitted that Excel is familiar to all of us, but that the consequence of its widespread use is decentralized and often incompatible R&D data. When a performance anomaly occurs – for example, one batch of cells performs poorly – it’s very difficult to track down whether the root cause was a particular lot of raw material or a different procedural step that is buried in a team member’s notes.
Instead, Moritz advocated for a centralized approach: a unified R&D platform like Albert that not only integrates ELN, LIMS, and inventory, but also layers AI on top of this structured data foundation.
.jpeg)
Optimizing electrolyte formulations with AI-ready data
To demonstrate what’s possible with the right foundation, Moritz presented a case study for electrolyte formulation. He began by collecting literature data from 19 papers, each of which reported ionic conductivity at various temperatures using different solvents and salts.
“Note that I did not have to write this data into Albert manually,” Moritz explained. “Once I had extracted structured information from the papers, I uploaded it in bulk using the Python interface of Albert. If I had performed the measurements myself, I could also have employed Albert Sync to directly parse information from a measurement device or network drive.”
After uploading this data into Albert, the scattered information – collected at different conditions and presented in different formats – became standardized, structured, and AI-ready. In this format, the data immediately began to generate useful trends.
“Once you put in the work of inputting your data in a structured way,” explained Moritz, “It becomes easy to pull out insights and trends.”
Once you put in the work of inputting your data in a structured way, it becomes easy to pull out insights and trends.
Unlike standalone AI solutions that require manual restructuring of files or scripting custom cleaning routines, the beauty of a unified R&D platform is that your data can be used for advanced analytics right away. Even without AI, Moritz showed examples of how scientists can quickly visualize ingredient usage trends, typical concentration ranges, or performance relationships across parameters like lithium salt content.
But the real power comes from Albert Breakthrough, our suite of AI/ML tools trained on over 15 million molecules. With Breakthrough, scientists can:
• Define target properties, such as conductivity
• Select constraints, such as requiring a specific salt, maintaining a certain ingredient ratio, or setting a cost limit
• View ranked candidate formulations with predicted performance, model confidence, and feature importance – including chemical properties from CAS data
In Moritz’s example, he constrained the model to use lithium hexafluorophosphate, but gave it considerable freedom in choosing solvent composition around that. The candidates suggested by Breakthrough can then be tested, and their results can subsequently be fed back into Breakthrough.
“Notably, the purpose of Breakthrough is not to bypass the input of an experienced chemist or formulator,” Moritz elaborated. “Breakthrough can make it easy for you to handle the complexity of optimizations along many parameters at once – that’s the part of research that human intuition does not handle well. However, the scientist defines the direction and limitations of the model based on her or his domain knowledge.”
Towards a cleaner energy future
From stationary energy storage to electromobility, the battery industry is making tremendous strides, and we look forward to keeping up with the latest findings at The Battery Show North America and the Advanced Automotive Battery Conference later this year. In the meantime, reach out or Request a Demo to learn how Albert can accelerate your battery R&D.
We recently joined over 1,000 exhibitors and 20,000 attendees at The Battery Show Europe in Stuttgart, Germany. As we connected with the brightest minds in the battery industry, a shared theme emerged: developing battery materials is inherently complex and entails large quantities of data. Moritz Haus, PhD, Solution Engineer at Albert Invent, spoke with Charged Electric Vehicle Magazine about how AI trained like a chemist can harness complex data to accelerate battery R&D:
Transforming battery complexity into battery clarity
Moritz took the stage and dove deeper into this topic through his presentation “Data Management and AI for Next-Generation Battery Materials Development.” He opened with an overview of a typical battery research workflow, revealing the reality that battery development is not just multi-step, but also multivariate.
“With so many factors at play, it becomes incredibly challenging to trace performance back to your formulation or process variables,” said Moritz. “Especially when that data is recorded across dozens of spreadsheets, in different formats, by different people.”
Moritz admitted that Excel is familiar to all of us, but that the consequence of its widespread use is decentralized and often incompatible R&D data. When a performance anomaly occurs – for example, one batch of cells performs poorly – it’s very difficult to track down whether the root cause was a particular lot of raw material or a different procedural step that is buried in a team member’s notes.
Instead, Moritz advocated for a centralized approach: a unified R&D platform like Albert that not only integrates ELN, LIMS, and inventory, but also layers AI on top of this structured data foundation.
Optimizing electrolyte formulations with AI-ready data
To demonstrate what’s possible with the right foundation, Moritz presented a case study for electrolyte formulation. He began by collecting literature data from 19 papers, each of which reported ionic conductivity at various temperatures using different solvents and salts.
“Note that I did not have to write this data into Albert manually,” Moritz explained. “Once I had extracted structured information from the papers, I uploaded it in bulk using the Python interface of Albert. If I had performed the measurements myself, I could also have employed Albert Sync to directly parse information from a measurement device or network drive.”
After uploading this data into Albert, the scattered information – collected at different conditions and presented in different formats – became standardized, structured, and AI-ready. In this format, the data immediately began to generate useful trends.
“Once you put in the work of inputting your data in a structured way,” explained Moritz, “It becomes easy to pull out insights and trends.”
Once you put in the work of inputting your data in a structured way, it becomes easy to pull out insights and trends.
Unlike standalone AI solutions that require manual restructuring of files or scripting custom cleaning routines, the beauty of a unified R&D platform is that your data can be used for advanced analytics right away. Even without AI, Moritz showed examples of how scientists can quickly visualize ingredient usage trends, typical concentration ranges, or performance relationships across parameters like lithium salt content.
But the real power comes from Albert Breakthrough, our suite of AI/ML tools trained on over 15 million molecules. With Breakthrough, scientists can:
• Define target properties, such as conductivity
• Select constraints, such as requiring a specific salt, maintaining a certain ingredient ratio, or setting a cost limit
• View ranked candidate formulations with predicted performance, model confidence, and feature importance – including chemical properties from CAS data
In Moritz’s example, he constrained the model to use lithium hexafluorophosphate, but gave it considerable freedom in choosing solvent composition around that. The candidates suggested by Breakthrough can then be tested, and their results can subsequently be fed back into Breakthrough.
“Notably, the purpose of Breakthrough is not to bypass the input of an experienced chemist or formulator,” Moritz elaborated. “Breakthrough can make it easy for you to handle the complexity of optimizations along many parameters at once – that’s the part of research that human intuition does not handle well. However, the scientist defines the direction and limitations of the model based on her or his domain knowledge.”
Towards a cleaner energy future
From stationary energy storage to electromobility, the battery industry is making tremendous strides, and we look forward to keeping up with the latest findings at The Battery Show North America and the Advanced Automotive Battery Conference later this year. In the meantime, reach out or Request a Demo to learn how Albert can accelerate your battery R&D.