CONTEXT Driven by the pedagogical transition to online delivery and assessment due to the COVID-19 pandemic, we have developed and implemented various online assignments and mastery quizzes using a versatile maths e-assessment tool called Numbas, for three mathematically intensive engineering courses (Mechanics of Materials 1 (2nd year), Structural Analysis (3rd year) and Control Systems (3rd year)). Issues in these courses included limited problem sets and student rote learning without understanding, assessments which had poor correlation to overall learning, and academic integrity concerns. We therefore aimed to use Numbas to improve the evaluation of students’ true understanding of the concepts, analysis procedures and problem-solving skills, thereby enhancing their independent learning experience and outcomes whilst minimising plagiarism. PURPOSE OR GOAL This study is to reflect on our three-year practice, evaluating the following: (1) How the online assessment using Numbas helped the students to consolidate their understanding in an effective way. (2) How breaking down large questions into consecutive and interconnected sub-questions helped the students to learn in a step-by-step manner. (3) How these online assessment helps to improve students reflective learning experience. APPROACH OR METHODOLOGY/METHODS Different methodologies were adopted for different courses, based on the course type and assessment type using Numbas. For structural and mechanics courses, comparative studies were performed on the correlation between the assignment and final exam performances between years with traditional paper-based versus e-assignments. For the control course, final exam performance was evaluated. Surveys were also conducted to provide evidence of how the online assessment has helped the students to learn. ACTUAL OR ANTICIPATED OUTCOMES Online assessments developed based on Numbas have demonstrated their effectiveness in enhancing the student learning experience and performance through well-structured formative e-assessment throughout the trimester. Featuring a large question bank with randomised questions, these e-assessments have served intended purpose of fostering independent learning, minimising plagiarism, and providing ongoing feedback to improve academic teaching and student learning. CONCLUSIONS/RECOMMENDATIONS/SUMMARY Purposefully designed and well-structured online assessments using Numbas have been developed and implemented for three mathematically intensive engineering courses at Griffith University. Findings suggest that such a development has resulted in positive impact to students learning experience and outcomes, while facing the unprecedented challenges of COVID-19.