Financial(AI)d

Inspiration

Since the onset of the COVID-19 pandemic four years ago, many people have been malaffected; including financially. However it taught people the possibility of long-term investing, especially with the rapidly growing stock and crypto markets. This phenomenon came with a downside, however, when people started blindly investing by "hype" or whatever is popular rather than putting effort into educating themselves. Hence, many have been vulnerable to market volatility and have lost much of their life savings.

Friends, families, and even myself lost hundreds, if not thousands, of hard-earned money since the downfall of Bitcoin, semiconductor index funds, and energy portfolios.

Applications such as Robinhood have democratized access to the stock market, making investing wildly popular among amateurs and experts. However, the challenging nature of investing can push many newcomers away with bad experiences, with volatile options like crypto and NFTs being a common talking point for many of them. We wanted to provide an interactive and intuitive means for newcomers to learn how to invest in more traditional sectors of the stock market in an interesting, safe, and informed manner.

As a team of amateur finance enthusiasts with very little experience, we found it can be hard for beginners to enter the world of finance without some (AI)d. Our project aims to provide an educational resource for newcomers and those interested in investing as a hobby.

What it does

Financial(AI)d draws on a large quantity of current news and data and feeds it to an LLM that makes informed, complex decisions to answer the user's query. The user can prompt a friendly and intuitive chatbot for a conversational-style learning-experience. The LLM draws on its dataset to inform the user of options and answer the client's questions.

How we built it

Our project runs as a Flask app in the back-end (connected with ChromaDB), served on the front-end by NextJS. For the LLM we utilise Google's Gemini 1.5 pro to make complex and informed decisions. We leveraged chain of thought prompting to reduce the chance of hallucinations and encourage the LLM to provide reasoned and informed responses.

Challenges we ran into

A major challenge was consolidating our disparate resources into compatibility with one another, for example the API with the LLM, the Chromadb with the LLM, the entire back-end with the front-end chatbot. Another challenge was learning many all new technologies such as Chromadb and Google's Gemini API.

Accomplishments that we're proud of

We are proud to leverage Chromadb's metadata filtering and low latency inserts/fetches. We are also proud of the information we were capable of gathering and integrating onto the database to further inform our LLM's decisions. Overall, we are excited to present a complex and nuanced chat-bot using many cutting-edge technologies in AI.

What we learned

We learned to leverage new and exciting technologies in AI to create a product that informs and assists the user. We learned a lot about data management and front-end development.

What's next for Financial(AI)d

Potential improvements include even more diverse database entries, better support for short-term investment advice, and always better data gathering. We would also like to integrate more AI machine-learning solutions to analyse raw data as opposed to string/JSON representations of articles.