Introduction: The majority of stroke centers do not have access to advanced imaging when treating acute ischemic stroke (AIS) patients with large vessel occlusion (LVO). We previously demonstrated that a machine learning (ML) algorithm could accurately predict infarct core from CT angiography (CTA). Here, we determine the algorithm's performance to predict clinical outcomes after endovascular therapy (EVT) compared to advanced imaging. Methods: From our prospectively maintained multi-center registry, we identified consecutive LVO AIS patients screened with CTA and treated with EVT. A deep neural network ML model was trained and validated to predict infarct core size from CTA, and 24-dimensional vector embedding was extracted for each patient. The primary outcome was performance of predicting 90-day functional independence (mRS 0-3) and was determined by area under the ROC curve for two classifier models (k-nearest neighbors, k=11) that included clinical data (age, gender, NIHSS, TICI) and either the ML embeddings from CTA (KNN-CTA) or CT Perfusion infarct core estimations (RAPID, iSchemaView) (KNN-CTP). Results: Among 119 LVO AIS patients who underwent EVT, median age was 68 [IQR 57-78], 45% were female, median NIHSS was 16 [IQR 13-21], and median ASPECTS was 7 [IQR 6-8]. Average CTP-RAPID core volume was 17.9mL ±27.5 (mean±SD). 94.1% of patients achieved TICI 2b-3 and 52.1% achieved mRS 0-3 at 90 days (Table 1). Point estimates for AUROC were superior for the KNN-CTA classifier compared to the KNN-CTP (0.72 vs 0.63, p=0.025) as shown in Figures 1 and 2. Conclusion: A ML model analyzing more widely available CTA images demonstrated superior results compared to industry standard advanced imaging in predicting clinical outcomes for LVO AIS patients after EVT. These findings suggest ML-based approaches may reduce the need for advanced imaging in EVT clinical decision-making.