In 2020, I created QuickFin.AI. Users could type questions about companies in plain English (e.g. "how were google's last year sales" or "why did estee lauder oper margin increase in 2019?") and get results almost instantly. Behind the scenes I built a pipeline of neural nets which would search 17,000+ plain text SEC filings and 2.3 million data-points to fetch the answer. QuickFin was featured on ProductHunt.
The big picture idea was to make the investment analysis process faster and friendlier. The pro investors could find answers to their questions much faster than a manual search, and could spend the bulk of their time making judgement calls. The budding investors would be able to comfortably interact with financial data using QuickFin's natural language interface.