Inspiration
Tired of stale news and empty chatter? Curious about any companies to invest in? Have any questions before diving into the exciting world of finance? Third Street Journal has you covered. We are your one stop destination for news, tips and insights.
What it does
Third Street Journal provides three main features: News: A culmination of hundreds of news articles compiled from multiple sources. These articles are also summarized by AI, and are a quick, efficient way to get a daily digest of fintech news. Stock Prices: A beautiful interface that allows users to select any public company and displays the opening, high, low and closing prices of the stock over a selected period of time. Chat: Our personalized, in house trained AI that aims to give you well informed guidance to explore the financial world and make informed decisions in markets.
How we built it
News: We built our network of news articles using a unique scraping system using proxies and user agents to bypass standard web scraping blockages, ensuring access to a wide range of financial content. We summarize this content using AI and vectorized this data to optimize for management and analysis. This data is stored in Pinecone, a specialized database for vector search.
Stocks: Our stocks interface was created by appending a Trading view advanced chart- a dynamically created script element, effectively embedding the chart into the site components.
Chat: We built this versatile agent that features real time stock market data from Yahoo finance with a base on openai's api. Outside of this, we used the yfinance API to train the AI with real time data. Additionally, we included a customized stock screener for targeted investment strategies and a semantic search for the processed articles in our data base to provide deep, contextually relevant insights.
Challenges we ran into
Initiating real-time communication within our chat interface presented a challenge. Establishing a stable connection concurrently with bootstrapping the chat state proved to be complex. Maintaining a seamless user experience while rendering a dynamic feed of messages took a lot of effort. Connecting the vector database and our AI API to the backend was a very slow process. We reduced the latency by optimizing by 60% by modularizing code for calling from the database, and running queries after setting the environment.
Accomplishments that we're proud of
Our AI delves deep into financial data and market sentiment. This multifaceted training fuels superior insights, not just predictions. GPT-4 focuses on language, but ours dives into financial nuance, crafting personalized, explainable guidance beyond competitors. It's a financial orchestra, not just a chatty chatbot. Our interactive stocks interface allows for easy access to any stock price and it displays all the data in a neatly formatted way. The news articles that we have compiled are expansive and high in number, giving us a large dataset to train our AI.
What we learned
What's next for Third Street Journal
We can further improve our AI by training it with more data. With access to valuable sources of data like Bloomberg and Wall Street Journal, we can provide users with the latest, most informative news articles and prepare them for financial decisions. We don't have the resources to scale to a large user base, or to store big data to provide real time services for our products with lower latency.
Log in or sign up for Devpost to join the conversation.