MSc Artificial Intelligence, University of Edinburgh
Krishnasree Kelath
I work on interpretable AI: models whose reasoning can be read, tested and compared with how human minds work.
Interpretability in machine learning. Cognitive science. Neuroscience. Theory of Mind.
About
I grew up in Dubai in a family of Carnatic musicians, studied physics, chemistry, maths and biology at school, and came to computer science sideways. In July 2026 I finished an MEng in Computer Science and Software Engineering at the University of Birmingham with First Class Honours: three years at the Dubai campus, then a final year at Edgbaston.
I'm now at the University of Edinburgh's School of Informatics for an MSc in Artificial Intelligence. I care about interpretability: understanding why a model gives the answer it does, and whether its reasoning looks anything like ours. That pulls me towards cognitive science and neuroscience as much as machine learning. Longer term I want to do doctoral research on cognitively informed AI and Theory of Mind.
Outside research I co-founded an AR travel app, led creative and operations at an indie game studio, and ran a public workshop on whether AI can be creative.
- Studying
- MSc Artificial Intelligence, University of Edinburgh, from September 2026
- Degree
- MEng Computer Science and Software Engineering, First Class Honours, University of Birmingham
- Research interests
- Interpretable AI, cognitive science, neuroscience, Theory of Mind
- Music
- Carnatic tradition; Trinity College London performance exams with distinction
- Based in
- Edinburgh
Research
I want models we can open up and question, and I use human cognition as the yardstick: if a model's reasoning can be laid out in terms people use, we can check it against how people actually think. My dissertation tested that idea on music, a domain I know from the inside.
Projects
Research, apps, talks, a satellite and some games. Open one for the details.
Experience
Recognition
Awards
Certifications
Machine learning
PyTorch, scikit-learn, librosa, torchcrepe, SVMs, LSTMs, CNNs, signal processing
Research
Interpretable ML, cognitive modelling, music information retrieval, cross-validation design
Languages
Python, Java, C, C#, JavaScript, HTML and CSS, LaTeX
Tools
Git, Unity, Blender, Figma, Vercel, Linux and Bash
Let's talk about how models and minds reason.
If you're working on interpretability, cognitive modelling or computational neuroscience, I'd like to hear about it.