FinanceJockey

Inspiration

The idea for FinanceJockey was inspired by the challenge everyday investors face when trying to make sense of overwhelming financial data — dense annual reports, volatile stock prices, and the constant stream of financial news.

We asked ourselves: What if there was a smarter way to combine all these signals into one intelligent assistant? One that not only reads reports and news like a financial analyst but also thinks like a strategist — providing personalized insights and actionable trading recommendations in real time.

We were particularly motivated by: -The rise of Retrieval-Augmented Generation (RAG) models and their potential to unlock insights from long, complex documents. -The gap between raw financial data and real-world investment decisions. -The vision of making financial literacy, analysis, and forecasting more accessible through conversational AI and intuitive design. -With the rapid growth of LLMs, sentiment analysis tools, and real-time APIs, we saw the perfect opportunity to build something that could change how people interact with the stock market — not just by informing them, but by helping them think, simulate, and act like professionals.

What it does

FinanceJockey is an advanced AI-powered financial assistant that empowers users to make informed investment and trading decisions by integrating company annual reports, live news, and real-time stock prices. Here are some of it's functions:

📄 Interprets Annual Reports: Uses Retrieval-Augmented Generation (RAG) to extract, understand, and summarize data from company annual reports — including text, tables, and financial metrics.

🧠 Smart Chatbot Interface: Allows users to interact with financial documents conversationally — ask questions like “What were Apple’s net earnings in 2023?” or “Compare R&D spending over the last two years.”

📈 Live Market Intelligence: Pulls in real-time stock prices and live financial news to build a comprehensive, up-to-date picture of a company’s market position.

📰 Sentiment Analysis on News: Analyzes tone and content of current news articles to predict short-term market sentiment, helping users understand if a stock is likely to move up or down.

🔮 Price Movement Predictions: Combines historical price data, company fundamentals, and news sentiment to forecast potential gains or losses.

💸 Actionable Trading Recommendations: Offers detailed buy/sell/hold suggestions with specific share amounts, entry points, price targets, and risk metrics, tailored to both short-term traders and long-term investors.

📊 Investor Strategy Guidance: Dynamically adapts its recommendations based on the user’s goal (e.g., growth vs. value investing), providing personalized insights for different investment horizons.

How I built it

It built by combining AI, finance, and real-time data pipelines into a seamless and interactive user experience. Here is the breakdown:

📄 PDF Parsing & Embedding -Extracted text, tables, and images from annual report PDFs. -Implemented a custom chunking strategy to prepare the documents for embedding using transformer-based models.

🔍 Retrieval-Augmented Generation (RAG) -Built a RAG pipeline where user queries are matched to the most relevant report content using vector similarity. -Integrated OpenAI’s GPT-4 as the language model to generate natural language answers grounded in the reports, news, and stock data.

📈 Real-Time Stock & News Integration -Pulled live stock prices and company fundamentals via financial APIs. -Integrated live news feeds using scraping and/or news APIs to monitor headlines, articles, and press releases in real-time. -Applied sentiment analysis models (VADER) to assess whether the news is bullish, bearish, or neutral.

🤖 Predictive Models & Recommendation Engine -Trained a custom financial forecasting model to predict stock movement using: -Annual report fundamentals -Price history -News sentiment trends -Built logic for short-term trading and long-term investment strategies, giving personalized recommendations on: -Share quantities -Entry/exit points -Expected gain/loss range

🧠 Frontend & UX -Developed with Next.js and Tailwind CSS for a fast, responsive UI. -Designed a clean and intuitive chatbot interface to let users ask questions or receive recommendations in plain English. -Enabled light mode theme using DaisyUI for better accessibility and readability.

Challenges I ran into

Some of the challenges our group ran into were: -Minimal resources on Indian stock market

integrating the Multi Model RAG with Next.js -Having the model accurately identify and analyze pictures

Accomplishments that I'm proud of

Some of the many accomplishments out team is proud of are: -Making 3 models within 24 hours -Utilizing an extremely large amount of API's and resources to bring the website together -Parallel running models -The UI looking clean -Staying up all night

What I learned

-Multi-model data fusion -Translating AI into trading strategy -Financial NLP and Sentiment modeling -A lot about stocks, markets, and annual reports

What's next for FinanceJockey

One of our main goals moving forward is to evolve FinanceJockey’s recommendation engine into a powerful simulation tool — where every stock suggestion isn’t just advice, but part of a predictive market scenario. By improving the accuracy and depth of our AI-driven recommendations, we’ll be able to simulate potential portfolio outcomes based on: -Hypothetical trades -Macroeconomic news trends -Long- and short-term strategies -This will help users visualize potential gains/losses, stress-test decisions, and become more confident investors.

We’re also taking our RAG model to the next level by making it truly multimodal — not just understanding text and tables, but also images, logos, charts, and branded visuals inside annual reports and company filings. This would allow FinanceJockey to: -Detect brand changes, partnership badges (e.g., SAP Gold Partner), and visual cues often overlooked in traditional NLP. -Extract insights from charts, graphs, infographics, and even certifications or award seals embedded in reports. -Merge that visual intelligence with financial data to create richer, more nuanced company profiles.