Machine learning researcher and engineer with experience in computer vision, time-series data analysis and forecasting, natural language processing, and graph ML. I am passionate about building reliable ML systems for real-world use and conducting research on new learning algorithms. My research interests include online learning and transformers for sequence modeling, video prediction, and generative AI. 🧑💻 🤖 I completed a Ph.D. at the University of Tokyo on respiratory motion forecasting with RNNs trained with online learning algorithms and transformers, and had the opportunity to investigate exciting physics and AI R&D problems across multiple industries (oil & gas, finance, healthcare, identity & security). I have been living in Japan for more than six years before moving to the UK. 🏯 💂♂️ I am happy to connect and exchange ideas with like-minded tech professionals as well as computer science and ML enthusiasts.
| Chest and liver cine-MRI prediction using PCA-based motion modeling and temporal dynamics forecasting with RNNs trained online and transformers | Deformable 3D image registration using the Lucas-Kanade pyramidal, iterative optical-flow algorithm |
|---|---|
|
Prediction of sagittal cine-MRI cross-sections 6 time steps in advance using sparse one-step approximation (left: ground truth, right: prediction).  |
Estimation of 3D lung-tumor motion due to breathing using deformable image registration.  |
| Time-series forecasting using online learning algorithms for RNNs and transformers | |
|
Prediction of the 3D positions of 3 markers placed on the chest of a subject 2.1s in advance (i.e., 7 time steps in the future, at a sampling rate of 3.33Hz) using RNNs trained with decoupled neural interfaces to guide the radiation beam during lung radiotherapy treatment.  |
|