Marcin is a mathematician and a Deep Learning computer scientist in one. An enthusiast with over five years of academic and commercial experience in DL/ML, Marcin specializes in Image Processing and NLP.
Marcin built a document classification system for predicting whether a given document is risky for an automatic translation – a project executed at Microsoft. The algorithm increased the accuracy of an automatic translation of Office 365 UI content by over 20% thanks to an effective detection of risky content.
At Sigmoidal, Marcin has led Object Detection and Activity Detection research. His research in Natural Language Processing allowed us to create a very strong article classification model, using a revolutionary diffusion technique – Positive Learning, which allows the model to be trained on a very limited amount of data.