BioMarket Tracker

AI-Powered Investment Analyst for the Biotech Sector

Comment

💡 Inspiration

As a freshman studying Finance and Bioengineering at HKUST, I noticed a massive gap in the market. Biotechnology stocks are notoriously difficult to analyze:

  1. High Barrier to Entry: Understanding clinical trial data requires Ph.D.-level science knowledge.
  2. Information Overload: Financial reports are hundreds of pages long.
  3. Language Barriers: Global investors often struggle with localized medical terminology.

I wanted to build a tool that democratizes this knowledge—an "AI Analyst" that works 24/7 to interpret complex bio-data for everyone.

🧬 What it does

BioMarket Tracker is an intelligent financial terminal tailored for the biotech sector.

  • Smart Ticker Search: Users can search by company name, ticker, or even Chinese Pinyin (e.g., input "maotai" to find Kweichow Moutai).
  • DeepSeek AI Analyst: Powered by the DeepSeek-V3 reasoning model, it generates institutional-grade investment memos. It doesn't just summarize; it reasons about risks (e.g., "Why did the Phase 3 trial fail?").
  • Real-Time Visualization: Interactive charts with technical indicators (RSI, MACD, Bollinger Bands).
  • Instant PDF Reports: One-click export to download a professional research report.

⚙️ How we built it

The project is a full-stack Python application:

  • Frontend: Built with Streamlit for a responsive, data-centric UI.
  • AI Core: Integrated DeepSeek API (V3). I engineered specific prompts to force the model to think like a Wall Street analyst (focusing on downside risks and catalysts).
  • Data Layer: Used yfinance for real-time market data and Plotly for interactive financial charting.
  • Logic: Implemented fuzzy matching for stock search to handle multi-language inputs.

🚧 Challenges I ran into

  • Prompt Engineering: Getting the AI to output structured Markdown instead of random chat text was tough. I had to iterate on the system prompts multiple times to ensure consistent report formats.
  • Data Latency: Fetching real-time data while generating AI responses caused UI lag. I optimized this by using Streamlit's session state to cache data.

🏆 Accomplishments that I'm proud of

  • Successfully integrating a Reasoning Model (DeepSeek) rather than a simple Chat model.
  • Building a complete SaaS-like experience (Search -> Analyze -> Export) in a short timeframe.
  • Applying my domain knowledge in Bioengineering to fine-tune the AI's focus on clinical trials.

🚀 What's next for BioMarket Tracker

  • Sentiment Analysis: Scraping news sites to add a "Market Sentiment Score".
  • Portfolio Management: Allowing users to simulate a biotech portfolio.
  • More LLM Models: Adding support for GPT-4o or Claude 3.5 as comparison options.

Built With

Share this project:

Updates