Fundamental questions in T cell differentiation remain unanswered: 1) EZH2’s role in Tfh fate commitment; 2) differences of TEX differentiation in Tumor and in Chronic infection; 3) CAR-T cells post-infusion differentiation. Besides, there is no high-dimensional quantitative analysis and machine learning pipeline for TCR signaling studies. This thesis work addressed these issues using advanced sequencing and imaging technologies. Firstly, using ATAC-seq, we found that EZH2-mediated H3K27me3 modifications regulate early Tfh fate commitment, but not late Tfh differentiation or memory maintenance. Secondly, through single-cell RNA- and ATAC-sequencing, we revealed that terminal effector-associated genes and loci accessibility were preferentially enriched in Tex cells in chronic viral infection, while stem/memory-related genes and their loci accessibility more enriched in Tex subsets in tumor, leading to differential responsiveness to PD-L1 immune checkpoint blockade (ICB). Thirdly, through longitudinal, single-cell multi-omic sequencing of YESCARTA CAR-T cells in patients with DLBCL, we characterized the heterogeneity of both CAR-T and endogenous T cells. We also found post infusion enrichment of effector memory/effector CD8+ CAR-T cells and regulatory CAR-T cells in responders and non-responders respectively. We also discovered an IRF7-centered regulatory module that can successfully predict clinical response. Furthermore, we developed an end-to-end pipeline that combines high resolution 4D LLSM data with machine learning and dimensionality reduction techniques to analyze TCR microcluster dynamics and predict T-cell signaling states. Taken together, this work demonstrates the power of single-cell sequencing and imaging technologies in enabling complex, dynamic, and deep-learning-based research on the molecular regulatory mechanisms and signaling patterns involved in T cell differentiation.