Abstract Speed of processing training (SOPT), practicing to enhance the information processing efficiency while performing various perceptual and cognitive tasks, is the widest examined type computerized cognitive training among aging populations, including those with mild cognitive impairment (MCI). However, the efficacy of existing SOPTs in maintaining or improving older adults' cognitive health greatly varies across individuals. Our preliminary studies identified that the flexibility of autonomic nervous system (ANS) is associated with learning and cognitive and neural gains in existing SOPT in older adults with MCI. Our premise is that adaptation capacity, which is primarily reflected by ANS flexibility, is a key contributor to the neuroplasticity underlying broad and sustained effects of cognitive interventions. In this proposed study, we will combine this ANS response profile with the traditional learning index to develop a “personalization engine”, called pSOPT, for better reflecting individual adaptation capacity, and to test pSOPT's feasibility (R21 phase) and preliminary effect (R33 phase) in MCI. In R21 phase (intervention refinement, Stage Ia), we will establish a “personalization engine” for the SOPT by taking advantage of unique information derived from ANS assessment that links to learning and test the feasibility in MCI. Advanced time-series data analysis methods (e.g., shapelet analysis) will be used to develop a prototype of pSOPT based on the previously identified ANS shapelet. Compliance and usability of the pSOPT will be examined using interviews, questionnaires, and recorded performance data using a single group design in older adults with MCI (n = 10). Specific aims include (1) Use the identified ANS shapelet to develop a “personalization engine” that can modulate SOPT according to real-time measures of ANS; (2) Examine the feasibility of administering the pSOPT. In R33 phase (pilot test, Stage Ib), we will test the preliminary effects of the pSOPT in MCI. A pilot double-blinded randomized controlled trial (RCT). An MCI group (N = 50) will be randomized into a 6-week pSOPT (n = 25), or attention control (n = 25). Cardiac monitor-based ANS signals will be recorded throughout training sessions across groups. Learning is indexed by performing accurately across a consecutive set of trials in training tasks. Cognitive battery (measuring cognitive gains) and BOLD fMRI-based brain function (measuring neural gains) will be assessed at baseline, post training (7-week), and short-term follow-up (3-month); neurodegeneration (T1MRI and blood- based Alzheimer's pathology) assessed at baseline. Specific aims include (1) Compare changes of cognitive and neural gains between groups; (2) Explore whether pSOPT will enhance ANS flexibility in supporting cognitive gains against baseline neurodegeneration. This study is a prerequisite to efficiently launch an efficacy trial of pSOPT in slowing dementia progress.