I like training all kinds of neural networks. Currently I'm working at PeakData AG with NLP and data processing and teaching and researchig at Wrocław University of Science and Technology as a part of cyber-physical-systems-group, where I work on teaching neural networks nonlinear dynamics.

Some things I'm currently working on:

• Libary for developing neural networks for nonlinear dynamics called pydentification
• Recreating transformer architecture on minimal examples for fun @ reconstructing-transformer
• Teaching university course about machine-vision, including processing, compression and neural networks
• Using AWS for machine learning and data processing with sagemaker and aws lambda

Some things I've worked in the past, which are left here:

• Developing neural networks with Fourier transform for nonlinear system identification, called frequency-supported-neural-network
• Using shallow neural networks with robust learning (changing the loss function) to constructing models of hardfacing process using experimental data @ hardfacing-neural-modelling)
• Using images of metal microstructure to estimate its hardness using fractal dimension computed from the image, visual-material-recognition
• Numerous time-series experiments with neural networks, including tensorflow-notebooks and university courses
• Some evolutionary-optimization and spea
• Some generative models and neural styling (with minal mobile-app)