Consolidating the key discoveries and learnings from the literature review and dataset evaluation, I compiled a problem discovery canvas to understand the current situation: large corpus, unstructured, complex, different styles of questions (keyword, longer natural), technical challenges

I then mapped out the problems to a solution process flow:

Finally, I translate the problems into a cohesive solution, with developer tooling alongside selected hardware, software, and cloud infrastructure

For much more detail, please see Chapter 4 of the Thesis paper