Machine learning as a tool for chemistry 

Recently, artificial intelligence has become a big talking point in scientific communities. However, the more efficient machine learning models are also the most complex, and the “black box” aspect to their functioning can discourage scientists from using them; especially in chemistry. One cannot afford to waste products or conduct dangerous experiments, so it makes sense to be distrustful of a machine’s “magic” predictions. Thus, my research internship focused on Explainable Artificial Intelligence (XAI), which as the name indicates, is artificial intelligence that explains to the user why it takes the decisions it does.​​

Several methods exist to force an artificial intelligence to explicit its decision making process, but I was encouraged to use Shapley Additive Explanations (SHAP). A simplified description of these explanations is that they calculate global weights of the parameters influencing the AI’s decisions. In chemical synthesis where many parameters have to be considered, SHAP are capable of explaining the predicted yield of an experiment as a function of these parameters. Using them, a chemist can easily make sense of the reaction scheme proposed by a machine.​

I exploited real life experimental results to train machine learning models and applied SHAP to analyze why they were or were not efficient. I then translated the graphical results into natural language, to have a ChatGPT-like experience with the prediction algorithm. There is more to this, so I encourage you to read the report on my profile! Additionally, do check out the publications from the Laboratory of Artificial Chemical Intelligence (LIAC) at EPFL who are doing amazing things on the subject.​

Artificial intelligence is a powerful tool that can quicken chemical research to a ridiculously high speed; in barely three or four years of concrete application, it has already won a Nobel prize. ​

My choice can appear random, especially given it was made in 2022 when artificial intelligence had just started emerging as a subject of interest in chemical research. Computational chemistry, or the branch of chemistry that happens in computers instead of laboratories, was and still is a domain predominantly focused on 3D modelization of compounds to better understand why they react the way they do and thus find new useful applications for them. This multidisciplinary subject interested me because it would not restrict me to a single branch of science. When presented with the many majors available in engineering, I wanted to study all of them—but that was obviously not possible, so I settled on chemistry thanks to the talented teachers who gave me a fascination for it. ​

On the side, I pursued personal coding projects encouraged by my father, a computer science engineer who had me discover the world of programming as a child. His research in machine learning is what made me look further into it, until I discovered that EPFL had a brand new computational chemistry laboratory mixing the chemistry I studied with my programming hobby. One year later, they accepted my request for a research internship and though they gave me directives and experiments to inspire me, they let me have free reign over my project. I am very lucky to have met the right people at the right time, and am excited to see how the future will pan out!​

I noticed very early during my leadership training that I really disliked asking for help. In fact, I spent my childhood being asked for help. Whether it be my younger sister for anything, my parents to translate administrative documents, or classmates who did not understand lectures; I was used to being the one people would come to when they needed something done, so I was too embarrassed to ask for anything back. The thing is, it is unrealistic for a single person to work completely on their own in an academic or professional environment, but for some reason I believed I could be the exception. Unfortunately, nobody can escape the burnout that comes when you isolate yourself with your issues. I still struggle with communicating to people when I need support, but the leadership training and experiences I have had during the Laidlaw program have definitely made me better at it.​

To be a Laidlaw scholar means having soft skills equally developed as technical skills. The leadership training we receive forces us, especially STEM students, to introspect and develop necessary social competences on top of our scientific knowledge to thrive in the professional world. ​

We can lose focus of why we study what we study, too busy with overwhelming lectures and deadlines, and the Laidlaw program teaches us how to step back and see the bigger picture that initially motivated us. Every scholar I have met has spoken about their projects with that look in their eyes that makes you excited to hear about it even if you had no knowledge nor interest in the subject before. Laidlaw scholars are inspiring in how they are able to share with others the often very complex and technical areas they study. That is the ideal I strive towards.​

I tend to appreciate leaders you can relate to because it makes my aspirations seem realistic. I did not really have role models growing up in France, because nobody on TV looked like me at the time. Possibly the kind of boring but true answer here is that my parents inspired me because they were the only adults I knew who I could see myself becoming like. They left their home country and moved to a new place with a new language and new people who were not always welcoming, yet they were able to integrate themselves and build a future for my sister and I. ​

Today, I am lucky to be studying in a place where I am surrounded by young leaders: the peers I interact with every day. You would be amazed at how much the people around you have achieved and/or are aiming to achieve, and you usually just have to ask to learn about them! As a Laidlaw scholar, it is very easy to find a new inspiring leader by simply talking to another scholar.​

If I started listing things I want in an ideal future, I reckon I would need over a thousand pages, but making sure we have a future sounds like a good starting point. The more I learn about chemical industry, the more I become aware of how urgent it is to change our ways if we want to keep living on Earth. It has become impossible to ignore the natural disasters we are both the victims and perpetrators of, but it is also impossible to make all industries become sustainable immediately due mainly to financial restrictions. ​

The future I want to contribute to create is one where we can still enjoy the nature surrounding us without having to drastically change our daily life. Most of the sustainable technology needed for this future already exists, now we need to ensure its safety, optimize it, and encourage its use around the world.​

Seeing the sea in my hometown!