In this thesis two techniques, attenuated total reflection (ATR) mid-infrared (mid-IR) spectroscopy and ultrasonic particle manipulation, were successfully combined for the acquisition of mid-IR spectra of cells in suspension. A prototype ultrasound accessory was designed, realized and combined with a custom fiber optic ATR probe. The working principle was successfully tested off-line with suspensions of Saccharomyces cerevisiae. A second prototype fit for use in a semi-industrial bioprocess environment was designed and built. It comprised the optimized ultrasound accessory and a commercial process spectrometer equipped with an in-line fiber optic ATR probe. Using this instrumental development, to the best of the author-s knowledge, mid-IR spectra of cells inside a stirred, aerated bioreactor could be successfully recorded for the first time. Changes of the carbohydrate content of S. cerevisiae undergoing nitrogen-limited growth, i.e. the accumulation of trehalose and glycogen, could be followed quantitatively off-line. Mid-IR spectra of washed, dried cells were recorded and analyzed by partial least squares regression (PLS-R) with reference values obtained by wet chemistry and liquid chromatography (root-mean-square error of cross validation: 0.33 % dry weight (%DW) for trehalose, 0.55 %DW for glycogen, and 1.17 %DW for mannan). Similar spectral changes could also be observed in-line using the novel ultrasound enhanced fiber optic ATR probe and principal components analysis of spectra of cells present in the fermentation broth. Quantification of the analytes proved to be difficult because changes in biomass and in the carbohydrate content occurred simultaneously, and this correlation could not be broken. Complimentary, the applicability of state-of-the-art mid-IR technology for quantitative analysis of solutes in-line was explored. On the example of Penicillium chrysogenum fermentations, PLS-R and multivariate curve resolution-alternating least squares (MCR-ALS) were applied for the simultaneous determination of penicillin V and phenoxyacetic acid. The obtained results show that models are of comparable quality with LODs in the low gL-1 range. MCR-ALS requires more process knowledge than PLS-R, but provides an objective assessment for model validity by comparison of the calculated pure component with the respective reference spectra. Using the same biological system, a fiber optic and a conduit ATR probe were simultaneously applied in-line. No significant difference in performance for quantification of the two analytes by PLS-R could be found. However, higher spectrum-to-spectrum stability and better signal-to-noise were found for the optical conduit set-up.