We welcome and congratulate Joel Burman on his new position as data scientist at SPKI. With us, he will get involved with a wide range of AI projects, including the development of the Quake II EPJ-dataset and working with machine learning algorithms developed based on time series data collected from, for example, ECG and vital signs at the intensive care units at UNN and Nordlandssykehuset.
Joel did his Master’s degree in the Machine Learning group at UiT. There he focused on few-shot image segmentation, making a segmentation map with only 1-5 images from the target class. He introduced a fine-tuning step in the inference phase for a previously well-performing network. This improved the results, especially in the previously low-scoring classes with a lot of in-class variation. This method can be applied to most few-shot segmentation networks. It could potentially also be used for health data when only small amounts of data are available.
During his studies, Joel also worked in a summer internship for the Machine Learning group. There he developed interactive networks for demonstration purposes using GAN (Generative adversarial networks) and text-to-speech to generate images och captions from a segmentation map drawn by the user.
In SPKI, Joel hopes to contribute to a more efficient healthcare system. Implementing Artificial Intelligence to repetitive time-consuming tasks he hopes that healthcare workers get more time for professional tasks. In addition, working on developing the Quake II database to help future research projects have clean and anonymous health data to work with.