The discovery of Denisovans is one of the most exciting findings in human evolution in the past decade. The striking similarity between sequences of the high-altitude adaptation gene EPAS1 in modern Tibetans and Denisovans suggested adaptive introgression. However, the time and geographic ranges where the adaptive introgression happened remain unknown. This talk consists of three studies that are related to archaic adaptive introgression. First, I show that modern Tibetans experienced two pulses of Denisovan introgression, among which a group more alike the Denisovans from the Altai mountains introduced the adaptive EPAS1 haplotype, and the positive selection on EPAS1 did not start until well-after the Last Glacial Maximum. Second, resolving the timeline of Denisovan adaptive introgression spurred an opportunity to reexamine Tibetan population history, which remains perplexing after decades of work. By leveraging genetic and archaeological evidence, I show that there are two possible models for the population history on Tibetan Plateau. Lastly, I introduce MaLAdapt – a machine learning method for detecting genome-wide adaptive introgression, which reveals novel knowledge of how genomic variants from archaic humans facilitated modern human adaptations in worldwide populations.