Financial Analysis Assistant

Financial Analysis Assistant is a state-of-the-art interactive AI assistant that provides objective and accurate financial data on public companies to assist you in your investment decision making.

Comment

Inspiration

An objective financial analysis tool to provide fundamental up-to-date financial data of any public company to the user and be able to answer financial questions about that company.

What it does

The Financial Analysis Assistant or FAA allows the user to provide a company name and quickly learn about the company, its management, key financial metrics, latest stock performance, and relevant legal considerations. The user can also chat with FAA and ask pertinent financial questions about investing in the company.

How we built it

First, we had to learn about correct prompt engineering to ensure we are asking the model questions with as much detail as possible, giving the model time to think and come to correct conclusion and ensure that the model does not hallucinate by instructing it to find relevant information before answering questions, the latter being particularly important for chatbot widget. Second, using the principles above, we formulated the questions and ensured that the answers are correct by running the model with multiple company names. Several trial-and-error runs helped us improve the instructions and questions to the model (iterative prompt development process). Third, we ensured that the widgets are aligned correctly with pleasing and easy to use flow for the user.

Challenges we ran into

Prompt engineering: Giving the model the right set of instructions to formulate answers is a challenging effort. E.g. the stable diffusion XL, logo generator does not properly work for all companies. It seems that the model is too creative when generating company logos.

Accomplishments that we're proud of

We are humbled at the amount of work required to create such sophisticated models in AWS. It is hard work to ensure models work correctly and ethically. We are proud that we adhered to these ethical standards and focused on data accuracy.

What we learned

Prompt engineering and its implications, as well as different Bedrock model types like: Claude, Jurassic, etc. by consulting the Bedrock site: link

What's next for Financial Analysis Assistant

Current FAA is the no-code proof-of-concept steppingstone for the ultimate project where the stock analysts can use to quickly decipher company CEO and senior management comments in real-time and formulate meaningful and relevant questions during quarterly earnings calls that would refer back to a specific past comment or financial data. The financial data that FAA would be searching through is stored as unstructured data like documents, podcasts and webcasts in public domains like SEC or company sites. The tool would ultimately show the financial data in graphical format and be implemented on AWS Bedrock and deployed to XR headsets like Meta Quest or Apple Vision Pro.

Built With

Share this project:

Updates