Tuberculosis (TB) is the world’s deadliest infectious disease, with over 1.5 million deathsand 10 million new cases reported anually. The causative organism Mycobacterium tuber-culosis (Mtb) can take nearly 40 d to culture, a required step to determine the pathogen’santibiotic susceptibility. Both rapid identification and rapid antibiotic susceptibility testingof Mtb are essential for effective patient treatment and combating antimicrobial resist-ance. Here, we demonstrate a rapid, culture-free, and antibiotic incubation-free drugsusceptibility test for TB using Raman spectroscopy and machine learning. We collectfew-to-single-cell Raman spectra from over 25,000 cells of the Mtb complex strain BacillusCalmette-Guérin (BCG) resistant to one of the four mainstay anti-TB drugs, isoniazid,rifampicin, moxifloxacin, and amikacin, as well as a pan-susceptible wildtype strain. Bytraining a neural network on this data, we classify the antibiotic resistance profile of eachstrain, both on dried samples and on patient sputum samples. On dried samples, we achieve>98% resistant versus susceptible classification accuracy across all five BCG strains. Inpatient sputum samples, we achieve ~79% average classification accuracy. We develop afeature recognition algorithm in order to verify that our machine learning model is usingbiologically relevant spectral features to assess the resistance profiles of our mycobacterialstrains. Finally, we demonstrate how this approach can be deployed in resource-limitedsettings by developing a low-cost, portable Raman microscope that costs <$5,000. Weshow how this instrument and our machine learning model enable combined microscopyand spectroscopy for accurate few-to-single-cell drug susceptibility testing of BCG.