BOLLaMa

• 

  BOLLaMa: Your AI sidekick for sustainable chemical innovation!

What it does

BOLLaMa, our AI-powered chatbot, simplifies the process of Bayesian Optimization for chemical reactions. It provides an easy-to-use interface that allows users to interact using natural language, removing the need for extensive programming knowledge or complex user interfaces. By offering a more approachable solution, BOLLaMa encourages the widespread adoption of sustainable optimization tools in the chemistry field.

How we built it

We built BOLLaMa using cutting-edge LLM (Large Language Model) technology to understand and process users' natural language inputs. We combined the LLM with advanced Bayesian Optimization algorithms to perform the optimization of chemical reactions. The chatbot's user interface was designed with simplicity and user-friendliness in mind, ensuring that even users with limited technical or chemistry knowledge can effectively engage with the tool.

Challenges we ran into

During the development process, we encountered several challenges. Adapting the Bayesian Optimization tools for real-world use cases proved difficult, as they were initially designed to showcase computational capabilities rather than user needs. Incorporating these tools within the LLM framework was also a complex task, requiring innovative strategies for seamless integration. Designing an interface that accommodates a diverse range of users, from novices to experts, demanded a careful balance between simplicity and functionality. Furthermore, creating an aesthetically pleasing and efficient UI involved the thoughtful combination of numerous tools, layout considerations, navigation elements, and visual design aspects.

Accomplishments that we're proud of

We're proud to have developed a user-friendly chatbot that democratizes access to advanced optimization techniques in sustainable chemistry. BOLLaMa not only simplifies the process for experienced users but also opens doors for those with limited technical knowledge to contribute to greener chemical processes. We believe that BOLLaMa has the potential to drive widespread adoption of sustainable optimization tools and make a tangible impact on EPFL's sustainability efforts.

What we learned

Throughout this project, we learned the importance of making complex tools accessible to a broader audience. We gained valuable insights into the challenges faced by the chemistry community in optimizing chemical reactions and the potential benefits of leveraging AI-driven solutions. Additionally, we learned how to effectively integrate LLM technology with advanced algorithms, which can be applied to a variety of other applications in the future.

What's next for BOLLaMa

We envision collaborations with labs at EPFL and industry partners to measure the tool's impact and gather valuable feedback for improvements. We aim to expand the coverage of chemical reactions, optimization objectives, and conditions, making BOLLaMa more versatile. To enrich the stand-alone usability, we'll incorporate additional tools that consider human safety, environmental risks, and sustainability issues of the proposed experiments. Lastly, we'll focus on enhancing explainability for the Bayesian Optimization process, fostering user trust and informed decision-making when implementing BOLLaMa's recommendations.

Built With

• botorch
• gauche
• gpt-4
• gradio
• langchain
• openapi
• pytorch-lightning