Staff Profile: Jaehoon Cha, AI Data Analysis Group Leader

Tell us about your role?

I am an AI Data Analysis Group Leader in Scientific Computing, working with multiple large-scale research facilities through the Ada Lovelace Centre. My role focuses on developing and applying state-of-the-art AI and machine learning methods to support scientific data analysis and facility operations.

My work covers areas such as scientific imaging, representation learning, surrogate modelling, optimisation, and AI-assisted workflows. I also collaborate closely with teams across facilities such as Diamond, ISIS and CLF, helping researchers explore how AI can be used in ways that are practical, reliable and scientifically meaningful.

What did you do before working at ALC?

Before working at ALC, I developed an academic background across mathematics, computer science and machine learning. I studied mathematics for both my bachelor’s and master’s degrees, with my master’s thesis focusing on homogenization analysis for nematic liquid crystals.

I then completed my PhD on artificial neural network design approaches for multi-channel information analysis, where I focused on developing neural network methods for understanding complex data. I also worked as a research fellow at the University of Basel on machine learning approaches for cryo-EM reconstruction, including methods for handling 2D projection data from unknown 3D molecular rotations.

What do you wish you had known earlier in your career?

I wish I had known earlier how important it is to work closely with scientists from different domains. Machine learning can support science in powerful ways, but only when we understand the scientific questions, the data, and the practical challenges behind them. For me, successful AI for science is not just about building good models, but about developing methods that genuinely help researchers make new discoveries.

What advice would you give to researchers who want to work in AI?

My advice would be to start with the problem, not the model. In AI for science, one of the most important steps is translating a scientific question into a clear and meaningful machine learning task.

It is also important to make sure the data is AI-ready. A large dataset is not enough on its own. Clear metadata, consistent structure and good documentation are essential for developing AI methods that are reliable, reusable and aligned with FAIR principles.

Finally, define what success looks like. A model should not only achieve good technical performance. It should be evaluated in a way that reflects the scientific question and the real decision it is intended to support.

Tell us something you have learnt about a project you have worked on recently

One thing I have learnt recently is that domain knowledge is essential when designing AI models for scientific data. If we want a model to extract meaningful features, we need to understand which variations in the data are scientifically important.

For example, in a cryo-EM project, centroid and orientation information were important for analysing different 2D projections. By designing an AI model that was aware of these variations, we were able to learn more meaningful representations and improve the way 2D projections were clustered.

What is your proudest career moment?

My proudest career moment is not a single event, but the career path I have built at STFC over the past five years. I have had the opportunity to deliver research outcomes while working closely with scientists across facilities such as Diamond Light Source, CLF and ISIS, and those collaborations have helped me grow as both a researcher and a leader.

Through this work, I have developed a clearer vision for how AI can support large-scale research facilities, which led to my current role as AI Data Analysis Group Leader. I am also proud that this journey has resulted in recognised research outputs and external recognition, including the Outstanding Research Award from the Korean Scientists and Engineers Association in the UK (KSEAUK).

What is the best thing about your role?

The best thing about my role is the opportunity to work with experts from many different scientific domains. Through ALC, I can collaborate with scientists across a wide range of facilities and research areas, learn new science, and see how AI can make a real impact in practice.

I find it especially enjoyable that every project brings a new scientific challenge. It allows me to keep learning while developing AI methods that can support real scientific discovery.

What one thing would you like people to know about your work at ALC?

One thing I would like people to know is that my work at ALC is about developing AI methods that can support multiple scientific domains, not just one. By working across different facilities and research areas, I have seen that fields such as astronomy, biology and material science often share similar challenges in extracting meaningful features from complex image data.

My goal is to design AI models that can learn useful and interpretable features from scientific images and be adapted across disciplines to support discovery. This has also shaped my recent work on semantic representation learning for scientific images, which was published in Nature Machine Intelligence.

What do you do when you are not at work?

Outside work, I enjoy running, hiking with good music, and spending time outdoors. The UK may not have many high mountains, but it has many beautiful hills, and I enjoy walking up them whenever I can.

I also love cooking, especially Korean food, and often invite friends over to share food together.

What is the best thing about working for ALC?

The best thing about working for ALC is the collaborative and supportive environment. ALC brings together expertise across AI, software engineering, data science and different areas of science, creating opportunities to learn from others and work on problems with real scientific impact.

I also value how ALC has supported my growth through cross-disciplinary projects, training and development opportunities. This has helped me broaden my scientific understanding and develop AI methods that can support wider research communities.

For more information about AI for Science: AI for Science | Machine Learning at STFC Facilities