Concept
Utilizing prompt interfaces and large language models to build a more robust and queryable financial application.
Longer Description
Utilizing natural language to both query simple data a la the original implementation of Kensho, as well as advanced data such as correlations, distributions over time, volatility levels, and much more.
Lastly also being able to utilize them to generate graphs and images related to the data you want to query “show me the quarterly net revenue of Uber versus Lyft” or “which insiders have sold more than $1M worth of shares over the past 3 months?”.
You even could enable very obscure data querys such as “what is the correlation between BTC price and home prices in the United States?”
Other thoughts on GTM/Founder Profiles etc
- There’s an interesting moat here in building all the piping to get all of this varied data. You’d effectively need a team that can do both BD + data scraping in addition to understanding the fine tuning of the model.
- I’m not entirely sure in what areas the real-time nature of the data would play a role in complexity.
- It’s likely this starts as a prosumer type tool and then would become robust enough once enough data integrations exist to go full enterprise.
- It’d be really fun if you could build in an execution layer over time “set up an iron condor for META for the April 14 expiry contract”. In addition to more sophisticated IFTTT integrations a la Composer
Comparable Companies
- Kensho
- Atom Finance
- Composer
- Adept.ai
- Robinhood
- Bloomberg
Related Reading
