Novel Spectroscopic Analysis for Breath-Based Diagnostics
The University is funded to develop a new type of spectroscopic tool that can deliver broadband and high-resolution spectra of complex molecular mixtures.
We are using this device to examine the exhaled breath so that we can potentially find evidence of undiagnosed disease based on the molecules that are present. The challenge of your project is to take these highly complex spectra, identify which molecules are present, and generate an estimate of the concentrations of each.
Unfortunately, under atmospheric conditions there are subtle interactions between the molecules that slightly change the characteristic spectral fingerprint of each molecule - this means simple nonlinear fitting or pattern matching algorithms do not work well.
Using a variety of machine learning and Bayesian inference algorithms, you will create better and faster ways to extract the important data from the observations.