What’s your name and where are you from?
My name is Usman Sani Dankoly, and I’m from Nigeria. I trained and worked as a physiotherapist there before my path in Artificial Intelligence in Healthcare and AI took me through Belgium, Germany, the Netherlands, and now Scotland. Nigeria is still “home” in how I think about community, health, and responsibility.
Where are you based and who are your supervisors?
I’m based at Glasgow Caledonian University, in the Data Science for Common Good group, working as a Marie Skłodowska–Curie Doctoral Fellow on the LABDA project.
My main supervisor is Professor Sebastien Chastin, and I’m co-supervised by Dr Philippa Dall, and Dr Zoë Tieges from Glasgow Caledonian University and Professor Jasper Schipperijn from University of Southern Denmark.
What’s your educational and professional background?
My background sits at the intersection of clinical practice, public health, and artificial intelligence.
- I started as a physiotherapist in Nigeria, working in hospital settings and medical rehabilitation care.
- I then moved to Belgium for a Master of Science in Rehabilitation Sciences and Physiotherapy at the University of Antwerp, where I graduated with distinction.
- I received the Belgian Government Master Mind Scholarship – a highly competitive international award – which funded my MSc and opened the door to my European academic journey. In parallel, I completed an Advanced Master in Public Health Methodology (Université Libre de Bruxelles), and later a Postgraduate Degree in Applied Artificial Intelligence (Erasmus University of Applied Sciences & Arts Brussels).
Professionally, I’ve worked as:
- A Project Manager and Data Scientist in a large international rehabilitation medicine hospital group in Berlin, developing patient-specific treatment pathways using causal machine learning and GANITE (a generative adversarial network for individual treatment effect estimation)
- An AI/NLP engineer, building pipelines for text classification, NER, topic analysis, sentiment analysis, and large-scale social media data.
- Most recently, I’ve specialised in Large Language Models—RAG systems, multi-agent architectures, vector search, and LLM evaluation and safety.
That mix of clinical practice, public health, and advanced AI is what I bring into LABDA.
What were you doing before joining LABDA, and how did you decide to join?
Before LABDA, I was working in Berlin as a project manager and data scientist, building machine learning systems to support clinical decision-making and personalise rehabilitation treatment. We were asking questions like:
- Which therapy combination is best for this specific patient?
- How long should we treat to get the optimal outcome?
It was powerful work, but I kept coming back to deeper questions:
- Who is actually represented in the data that trains these systems?
- Do our models behave differently for different social groups—even when we don’t explicitly model that?
- Are we quietly encoding inequalities into “smart” tools?
When I learned about LABDA, it felt like the right next step: a network that combines movement behaviour, intersectionality, and advanced data science, and is explicitly oriented toward the common good. It gave me the chance to bring my AI and clinical experience into a setting where we can rethink how we measure and model 24-hour physical behaviour in a way that is fairer and more inclusive.
What’s your PhD topic?
My PhD is about developing a universal taxonomy of 24-hour physical behaviour and behaviour dynamics, using narratives and large language models.
Instead of starting from accelerometer signals alone, I start from how people describe their day in words—through:
- 24-hour diaries,
- “Day in the life” vlogs,
- Social media narratives.
My project has four main components:
- αBET Framework – the backbone
I help finalise and operationalise the αBET framework, which models daily life as a sequence of discrete events—like “woke up”, “took the bus to work”, “looked after my mother”—each with a start time, duration, intensity, and context. This gives us a common language for describing behaviour across devices, studies, and countries.
- Intersectional sampling – who is in (and out of) the data?
Here I draw on a scoping review of intersectional sampling methods (for which I received the SPARC Best Student Contribution Award 2024) to think carefully about who we include when we build a “universal” taxonomy. I focus especially on language and geographic region as practical, ethical axes to ensure that people from very different backgrounds are represented in the diary corpus.
- “A Day on Earth” – building a global diary corpus
We are collecting single-day diaries from around the world, in many languages, and complementing them with publicly available “day in the life” vlogs. I then develop transformer-based Named Entity Recognition (NER) models to automatically detect events (e.g. “woke up at 6am”, “worked a night shift”, “prayed with my family”) and map them onto the αBET taxonomy.
- Predictive modelling of event sequences – understanding routines
Finally, I apply sequence-based transformer architectures (inspired by models like life2vec) to encode days as event sequences and study how routines evolve. The aim is to move beyond just “how many minutes of X” and towards understanding the structure, transitions, and stability of real-world 24-hour behaviours.
What would you like to achieve with your research?
I’d like my work to contribute to three things:
- A culturally inclusive, future-proof behavioural taxonomy that researchers and policymakers can use across countries, languages, and contexts—without forcing everything into a narrow, WEIRD template.
- Practical tools and methods so that this doesn’t remain a “nice theory”, but something people can actually apply to real-world data.
- A concrete demonstration that AI and large language models can be designed to recognise diversity instead of flattening it—and that this leads to better, fairer health insights.
If, a few years from now, researchers or policy makers from around the globe can use this taxonomy and it reflects their community – then the PhD has done its job.
What’s your role within LABDA?
Beyond the research, I also co-work as part of the Dissemination, Communication, and Events (DCE/DEC) Committee.
Can you tell us one personal thing about you that is weird or funny?
One funny thing about me is that the only meal I can confidently make really well is noodles.
