Epoch

Inspiration

Retail Trading has tripled in last last 5 years, driven by a massive influx of new retail traders. Companies like Robinhood, Fidelity, and Public have democratized access to trading for retail traders. However, access to financial data is still exclusive, expensive, and concentrated with a few. With Epoch, we want to change that by providing comprehensive and simplified access to financial data for everyone.

What it does

Epoch redefines the trading experience by granting both professional and retail traders unparalleled access to real-time public financial data across companies, equities, and options markets. It harnesses the power of large language models, providing traders with invaluable insights to inform and enhance their trading strategies.

How we built it

We created a function-tool agent that leverages several open APIs e.g. finnhub etc. to gather information for different varieties of queries. A real-time database is setup using AstraDB which monitors daily price and volume change, reported and expected earnings for SP500 companies. All these data sources are then leveraged by a hierarchical agent router backed by LLamIndex to figure out the best path to answer a user query.

Challenges we ran into

Financial data is massive and disorganized. From the beginning, our goal was to identify the most critical data sources for our target segment and build data pipelines for those datasets. Our biggest challenge was to bring all these different kinds of datasets such as stock prices (structured), earnings reports (unstructured), and news (real-time unstructured), that change on a regular cadence, together and build an intuitive experience on top of them.

Accomplishments that we're proud of

User-Centric Development: Successfully identified and prioritized pain points of potential users through extensive market research. Also, engaged with a select group of power users who are excited about the project, committed to trying out the product, and providing valuable feedback for iterative improvements.
Interest from Community: We have been able to garner some interest from academia and other researchers in the space, who are willing to collaborate in building state-of-the-art fintech-specific LLMs and Databases. This will help us build a product that is much more user-friendly and scalable.

What we learned

Financial data is massive, disorganized, and real-time. Firstly, it takes significant effort to bring them on a single platform. Furthermore, even though LLMs can help with accessing different types of data to gather useful insights, a chat interface might not be the best user interface for accessing this data. Based on our early user feedback, it is extremely critical to build a user interface custom-built for user trading and investment research workflows.

What's next for Epoch

As the next steps, we have the following tasks on our roadmap:

Add more data sources and capabilities: Currently, we have provided access to direct and basic financial data. As the next steps, we want to explore more complex data sets, including building our own knowledge graph to help answer more complex queries.
Alpha testing with users: As mentioned above, even though technology can help bridge the gap of accessing vast amounts of data quickly, we'll work extensively with users to understand how Epoch can be most effective in their workflows. We have identified 15-20 users with whom we will conduct alpha testing to collect feedback and iterate on it.