For use in my main projects within AI for microscopy and spectroscopy. In concrete terms, I utilize deep learning techniques in an optical microscopy setup to study the properties of nano-sized biological molecules, in a catalytic nanoreactor to study the effect of complex catalytic nano-particle surface effects on said particles' catalytic activity, and improve hydrogen gas sensing in palladium-based nanoplasmonic sensors. Results on simulated biomolecule data show that an order of magnitude improvement in limit of detection is possible when employing deep-learning based computer vision models compared to standard particle tracking algorithms. If this can be experimentally validated, it means that direct measurement of the size and mass of the protein hormone Insulin is possible for the first time; potentially opening up entirely new avenues of diabetes research and microscopy-based biological research in general. Results on experimentally measured catalytic activity show that direct measurement of a single catalytic nanoparticle is possible, corresponding to three orders of magnitude better limit of detection that current state-of-the-art. The approach in this project was inspired by the work done on denoising gravitational wave signals in noisy LIGO detector readout. Results on experimentally measured nanoplasmonic readout of hydrogen gas sensors show that a factor ~5 sensing acceleration is possible, as well as a ~2x improvement in limit of detection as compared to standard analysis methods through use of recurrent ensemble modelling techniques. All aforementioned projects are currently at different stages of being written into manuscripts, and SNIC resources would be vital to allow me to train all these models concurrently without unecessarily having to triage and prioritize certain projects.