Abstract: Currently time-domain astronomy can scan the entire sky on a daily basis, discovering thousands of interesting transients every night. Classifying the ever-increasing number of new transients is one of the main challenges for the astronomical community. One solution that addresses this issue is the robotically controlled Spectral Energy Distribution Machine (SEDM) which supports the Zwicky Transient Facility (ZTF). SEDM with its pipeline pysedm demonstrates that real-time robotic spectroscopic classification is feasible. In an effort to improve the quality of the current SEDM data, we present here two new modules, byecr and contsep. The first removes contamination from cosmic rays, and the second removes contamination from non-target light. These new modules are part of the automated pysedm pipeline and fully integrated with the whole process. Employing byecr and contsep modules together automatically extracts more spectra than the current pysedm pipeline. Using SNID classification results, the new modules show an improvement in the classification rate and accuracy of 2.8% and 1.7%, respectively, while the strength of the cross-correlation remains the same. Improvements to the SEDM astrometry would further boost the improvement of the contsep module. This kind of robotic follow-up with a fully automated pipeline has the potential to provide the spectroscopic classifications for the transients discovered by ZTF and also by the Rubin Observatory’s Legacy Survey of Space and Time.