Drill and practice is a well-received approach to repeatedly train learners' skills through a series of exercises and to reward them with corrective feedback. However, drill-based training may not improve learners' performance if its exercises are badly designed (e.g., not fun, not relevant to the learning goal, and becoming too difficult or too simple). To make drill-based training more effective, researchers have been designing two types of e-learning tools: educational video games to motivate learners to practice and intelligent tutoring systems to personalize the exercises. Nonetheless, the existing e-learning tools have two main problems: (1) not providing situational context for training, which limits learners' ability to apply previously learned knowledge and skills to real-life situations; (2) not designed for learners with specific learning disabilities (i.e., dyslexia, dyscalculia, and dysgraphia). To design educational video games that provide situational context for training, I sought to leverage entertainment technologies (e.g., storytelling technology) and AI technologies (e.g., computer vision technology). I tackled two specific projects. The first project was to improve anti-phishing training by simulating actual phishing attacks in a role-playing game. The key design challenge was to design an intensive workplace atmosphere that is easy for learners to make mistakes but also keeps the pleasure of completing tasks. My approach was to design an interactive storyline about building business as a banker and to design tragic endings that are realistic but with a sense of black humor. The second project was to teach vocabulary for objects located in the player’s immediate vicinity. The key design challenge was to guide learners to interact with learning materials in the physical world. To meet this challenge, I designed a new selection highlighting mechanics for AR scenes and designed an AR progress bar to indicate players' selection progress. Evaluations showed that my two educational video games for training were fun and also improved learners' post-test performance in practice. To help learners with SLDs use math e-learning tools for math skills training, I started by conducting an interview study with teachers for SLDs to study the difficulties that learners with SLDs faced in using e-learning tools. According to the interview study findings, learners with SLDs needed teachers to help them manage negative emotions during drills and practice. However, teachers were often unavailable during students' independent exercises. Therefore, I designed an intelligent tutoring system that detects and mitigates learners' negative emotional behaviors. This system analyzes eye-gazing data together with other traditional input data to detect learners' negative emotional behaviors. I also designed four intervention methods to mitigate the negative emotional behaviors: (1) praising for correct steps to solve the problem, (2) providing hints, (2) switching to a simpler problem, and (4) offering brain breaks. I conducted a formative study with teachers for SLDs to refine the design of the intelligent tutoring system. Teachers agreed that this system would help learners with SLDs reduce negative emotional behaviors. They also suggested that the system, in the future, should personalize the detection of negative emotional behaviors to help students who have more severe learning disabilities. Overall, my dissertation answered three main research questions: (1) how to design educational video games that provide situational context to practice skills; (2) whether and how learners with SLDs have difficulties in using the existing e-learning tools to practice math skills; and (3) how should we design the e-learning tool to intelligently tutor learners with SLDs. At the end of my dissertation, I discuss the limitations of my doctoral work and propose potential future directions for designing educational video games and intelligent tutoring systems for drill-based training.