MLIR is the driving force behind fast ML algorithm compilers as well as the high level synthesis tools targeting fast programmable logic or AMD AI Engines. This could enable running complex workloads on these devices from a high level ML-library. However, the MLIR tools surrounding these platforms however speak a different MLIR dialect than ML-libraries and most required (control-flow) operations are not supported. You task would be to build a translator between these different dialects, and to show how this would enable executing complex algorithms in new context (to be released in the MLIR ecosystem).
IMPORTANT: When applying for a thesis, always attach your CV, course list and course grades to your application so as to ensure best fit with the topic at hand!