<h3>Purpose</h3> Preoperative post-LVAD RVF predictive models have moderate discrimination. We hypothesize that machine learning using unsupervised k-type phenotyping can identify homogenous subclasses of patients in this heterogenous population. <h3>Methods</h3> A multi-center retrospective study of 826 consecutive patients who received either a HeartMate2 or HVAD between 1/2007 and 12/2016. After removing outliers, 816 patients were clustered using an unsupervised k-type approach with the following 10 variables: mean PA pressure (mPAP), hemoglobin (Hg), absolute neutrophil count (ANC), absolute lymphocyte count, CVP/PCWP ratio, age, INTERMACS profile, transpulmonary gradient, body mass index (BMI), and cardiac index (CI). Characteristics associated with cluster membership were assessed using a multinomial logistic regression. <h3>Results</h3> Three clusters were identified. Cluster 1 had a mean age of 53 years old, INTERMACS profile ≥ 3 (70%), borderline hemodynamics, normal labs and BMI. Factors associated with Cluster 2 membership included older age (RRR 1.09, 95% CI 1.03-1.16), elevated BMI (RRR 1.25, 95% CI 1.13-1.38), low INTERMACS profile ≥ 3 membership (RRR 0.04, 95% CI 0.01-0.21), low Hg (RRR 0.23, 95% CI 0.15-0.37), elevated ANC (RRR 2.07, 95% 1.36-3.14), high diastolic pulmonary gradient (DPG) (RRR 1.26, CI 1.10-1.45) and PVR (RR 2.06, CI 1.52-2.80). Cluster 3 membership was defined by old age (RRR 1.29, 95% CI 1.18-1.41), high ANC (RRR 2.45, 95% CI 1.57-3.83), normal CI (RRR 18, 95% CI 4.7-69.3), lower BMI (RRR 0.75, 95% CI 0.65-0.86) and hemoglobin (RRR 0.12, 95% 0.06-0.23). Cluster memberships stratified receipt of unplanned RVAD or death from RVF during index admission (Cluster 1: 5.1% [ref], Cluster 2: 17.4% [OR 8.85, 95% CI 12.14-7.38], p < 0.001], Cluster 3: 9.5% [OR 1.98, 95% CI 1.03-3.80, p = 0.04) and one-year survival (Figure 1). <h3>Conclusion</h3> We show three distinct patient phenotypes who develop post-LVAD RVF. Therapy targeting these patient groups could help mitigate RVF.