Typically, species tend to form when populations are isolated from breeding with each other. Unless the world becomes significantly less globalized, and remains so for millions of years, it is unlikely for humans to speciate through that route.
Pure gene therapy is unlikely to be enough. Today, most people see particular qualities as being "good" or "bad", and most genetic engineering would be focused on making people "better". This would likely result in decreased diversity among the human species, rather than the fragmentation you're looking for.
However, a societal reform, a change to the way we think of "individuals", along with gene therapy, could provide another route.
Imagine a world where every person is born into a particular job, where their entire worth is determined by how well they perform at that task. Families grow more suited for their task, and tend to produce children with others who have the same societal role. This could, theoretically, result in fragmentation, where each "breed" of human is built to their task - much like dog breeds have been historically. You could have bulky construction workers, autistic super-genius engineers, subterranean tunnel-workers with good night vision, hairless and adaptable space explorers, even cute and docile "human pets". Breeding humans who excel at and enjoy their assigned task may prove to be easier and more efficient than creating robots for the same purpose.
While it is unlikely that such a society could last long enough for true speciation to occur (even dogs, which have been bred for thousands of years, and have tremendous diversity in form, are still the same species), genetic engineering could accelerate the process. Given enough time, human society could wind up as an ecosystem, with each human species providing something of value to the whole.