Job's description:

Motivation: Nuclear magnetic resonance (NMR) has become indispensable in various research fields such as physics, chemistry and medicine. It provides a non-invasive and non-destructive method for examining a wide range of samples, for example, to find out the composition of SARS-CoV-2 at an atomic level.
The outcome of such NMR measurements is often a spectrum, which precisely describes the content of the examined sample. Deep learning methods are currently utilized to process the spectrum and extract features. However, the spectra are often of high dimensions, and thus, the computational demands for processing them are high.
Luckily, spectra contain unimportant information like noise, and a small set of variables can often describe the critical features.  

Your task:

You will use deep learning techniques to compress spectra into a set of descriptive features that are enough to reconstruct the spectrum from them. Specifically, autoencoders (or encoder-decoder networks) are a promising approach: The encoder compresses the input to a small latent representation, and the decoder can reconstruct the input given this latent space. Your task includes finding the most suitable autoencoder structure (e.g. VAE) to compress NMR spectra, and to implement a functioning pipeline in pytorch (or similar).
You will be part of Prof. Korvink’s research group where you can get support from members with expertise in NMR theory, methodology, hardware, and simulation.

Personal qualifications:

• Highly motivated student with excellent academic record
• Excellent knowledge of programming language python (or similar)
• Experience with deep learning algorithms and knowledge about state-of-the-art methods
• Optional: Basics in NMR
• Languages: English or German

click here for more information and Online apply

Technical contact