Aims: ECG-based risk stratification tools enable large scale screening and targeted prevention for heterogeneous conditions such as coronary artery disease (CAD). ECG-derived deep learning embeddings via trans-fer learning have demonstrated their utility in cardiovascular disease detection and prediction, but their utility in identifying cardiovascular risk trajectories remains unexplored. We aimed to extract latent features from a pre-trained model to refine CAD risk phenotypes identified in previous work and evaluate their clinical interpretability. Methods: Context latent features were extracted from 10-second, lead-I ECG recordings of 1,928 individuals with CAD from the UK Bi-obank using a Contrastive Predictive Coding model. The principal com-ponents (PCs) of these features were calculated and used as input to a k-means clustering model to determine distinct clusters in an unsupervised manner. Then, we analyzed the association between the identified clus-ters and incident outcomes (i.e., atrial fibrillation, AF, and heart failure, HF). Additionally, multivariable associations between the latent PCs and outcomes were analyzed, and Pearson correlations with traditional ECG parameters were performed to assess clinical interpretability. Results: The latent PCs identified two clusters of individuals with Cluster 2 having significant higher incidence of events for AF (8.9%) and HF (6.6%). Although the cluster-based risk stratification was less pro-nounced than that achieved with traditional ECG parameters, individual PCs demonstrated robust predictive ability for both outcomes. We identi-fied specific and shared PCs across outcomes, suggesting a common subclinical latent risk profile. Furthermore, the PCs most predictive of AF and HF correlated with markers of slowed ventricular conduction and prolonged repolarization, both established signatures of electrical re-modeling. Conclusions: Our findings confirm that latent features hold discrimi-native value for cardiovascular outcomes in CAD populations. While individual features provide risk information, future work should explore alternative clustering architectures to enhance the risk stratification us-ing latent features.