🧠 Brainiac: Agentic AI-driven Quantitative Alpha Builder

Welcome to Brainiac! Created in 2024, this project aims to streamline the generation, refinement, and evaluation of financial alphas.

Brainiac leverages agentic AI technologies to autonomously navigate through vast amounts of financial research, crafting and optimizing financial alphas. The goal is to equip quantitative analysts and traders with a powerful tool to streamline alpha generation.

🌟 Key Features

• Agentic AI System: Utilizes a cutting-edge AI framework for autonomous scraping and analysis of financial research papers, extracting actionable insights to craft financial alphas.
• LangChain Integration: Employs LangChain to foster an agentic reinforcement learning framework, enhancing the alpha generation process through continuous learning and adaptation.
• Alpha Parsing and Dataset Retrieval: Converts generated alphas into WorldQuant Fast Expression language, leveraging RAG (Retrieval-Augmented Generation) to find the right dataset tailored to each alpha.
• Automated Backtesting: Seamlessly integrates with WorldQuant's BRAIN platform, enabling automated backtesting and in-depth performance evaluation of the generated alphas.

🔧 Under the Hood

graph LR
A[Data Scraping and Analysis] --> B[Alpha Generation]
B --> C[Alpha Parsing]
C --> D[Dataset Retrieval]
D --> E[Backtesting]

Loading

1. Data Scraping and Analysis: Scrapes financial research papers from diverse sources, employing natural language processing techniques to sift through the content and unearth potential alphas.
2. Alpha Generation: Uses the LangChain framework to apply reinforcement learning for refining and optimizing the initial alphas, ensuring their viability and profitability.
3. Alpha Parsing: Transforms the refined alphas into WorldQuant's Fast Expression language, preparing them for dataset retrieval and backtesting.
4. Dataset Retrieval: Engages RAG to fetch the most relevant datasets for each alpha, ensuring that backtesting is conducted with the most suitable and comprehensive data.
5. Backtesting: Conducts backtesting of the alphas, now paired with their datasets, on the WorldQuant BRAIN platform, yielding detailed performance evaluations.

🚀 Getting Started

Before getting started, please ensure you have the following:

• Access to financial research papers and datasets.
• An account with the WorldQuant BRAIN platform for backtesting.
• A foundational understanding of financial alphas and quantitative trading strategies.

📥 Installation

Follow these steps to set up the project:

# Clone the repository
git clone https://github.com/jdhruv1503/Brainiac.git

# Navigate to the project directory
cd brainiac

# Install dependencies
pip install -r requirements.txt

After this, add your WorldQuant BRAIN credentials as in the format in example.json in the Simulation/creds folder, in multiple json files.

🔑 API Keys

Brainiac requires API keys from LlamaParse and DeepInfra to function properly. Follow these steps to set up the necessary API keys:

• Obtain API keys from LlamaParse and DeepInfra.
• Set the LLAMA_PARSE_API_KEYS environment variable to a comma-separated list of your LlamaParse API keys.
• Set the DEEPINFRA_API_KEYS environment variable to a comma-separated list of your DeepInfra API keys.

For example:

export LLAMA_PARSE_API_KEYS="your_llamaparse_key1,your_llamaparse_key2"
export DEEPINFRA_API_KEYS="your_deepinfra_key1,your_deepinfra_key2

🎯 Usage

# Run the simulation API
cd Simulation
python3 main.py

Now that this is running in the background, dump all your research PDFs into the AlphaGenerator/ResearchPDFs folder. Dump all text strategies into the AlphaGenerator/Strategies folder. Then, run python3 watchdog.py. This will look for changes in these 2 folders, when a new source is detected, will create alphas and send them to be backtested to the running simulation API.

📊 Results

Your alphas should be stored in a systematic way as a SQLite database Simulation/alphas/alpha_results.db

🔮 Future Enhancements

• Continous learning and alpha improvement agent framework to be added
• Automated scraping of strategies and SSRN papers
• Integration of USA datasets

🙏 Acknowledgements

Special thanks to the incredible RussellDash332/WQ-Brain repo, which served as the foundation for the API automation for BRAIN!

---

Start building powerful alphas with Brainiac today and revolutionize your quantitative trading strategies! 🚀