First Demo Video of the Application

Extract and Analyze workflows / Summary and Chat with Fundamentals

QuantCoder FS – Development Update (April 8)

This video presents the current state of development of QuantCoder FS, an AI-driven assistant designed to support research in quantitative finance.

The demonstration begins with the Summary workflow, which enables users to upload a PDF article and obtain a concise summary generated through natural language processing techniques and large language models. The output is produced in approximately 30 seconds and can be saved or copied for further use. A “Code” button, originally implemented to directly generate trading algorithms from summaries, is being repositioned within the interface due to architectural constraints.

The coding module remains under active development. A legacy version of the code generation engine is publicly available on GitHub. Results obtained from early iterations of this module—such as code compilation rates and example strategies—are discussed in articles published on Medium, with links available via the project's Substack page. A future LLM Inference panel is planned to enhance transparency by allowing users to trace the reasoning steps performed by each agent.

The Chat with Fundamentals workflow provides functionality for analyzing financial securities through the integration of EODHD data APIs. Users can retrieve and export results in CSV or JSON format. While the tool facilitates structured financial analysis, it is not intended to provide investment advice. Future iterations may include additional capabilities such as commodity price forecasting.

The platform is supported by a Substack publication offering an introductory article, direct access to the GitHub repository, and a high-level system overview. A comprehensive refactor is currently underway to improve the system’s scalability and modularity, emphasizing a plug-and-play design for individual workflows.

QuantCoder FS is steadily advancing toward becoming a robust platform for the integration of essential tools in quantitative research.

Disclaimer: Nothing presented in this video constitutes a recommendation to buy or sell any financial instrument. Trading and investing inherently involve risk. Viewers and users are encouraged to conduct their own due diligence.

Full transcript

Hello, my name is Emilia Bennett, and I’m the Eleven Labs voice chosen to present the current state of development of QuantCoder FS, as of April 8.

After logging in with a test user, we land on the homepage and start with the Summary workflow. We upload a PDF article and request a summary. In the backend, you can see the agents at work. The idea is to make this workflow domain-agnostic, using NLP techniques to clean and prepare the text before handing it to a large language model.

In about 30 seconds, we get a concise and insightful summary. Classic buttons allow saving or copying it. Originally, a “Code” button was added to directly generate a trading algorithm, but it will be moved to a higher-level access point due to architectural constraints.

We then visit the coding area, which is still under development. A legacy version of the engine is available on GitHub and was previously integrated into the app. While initial tests suggested 70% of generated code compiled successfully, actual performance may be lower. Still, the tool gained traction with around 150 clones per week and 50 stars on GitHub, showing growing interest in LLM-based pair coding for trading.

A future upgrade will introduce a dedicated LLM Inference panel to visualize the steps taken by each agent. This is essential in fields like quantitative finance, where users need to understand and retrace the logic behind the output.

Next, we explore the Chat with Fundamentals workflow. This lets you analyze or compare securities using financial data, news, and fundamentals from EODHD, a top-tier data provider. All data can be downloaded in CSV or JSON format, but usage remains subject to EODHD’s terms of service.

It’s important to note that this tool isn’t designed to provide direct investment advice. Instead, it’s meant to give analysts quick access to structured data, and help investors identify where to focus their attention. This pipeline could soon power additional tools, like commodity price forecasting.

Now, I’ll briefly show you the application’s dedicated Substack, which includes an introduction article explaining the project’s philosophy and development roadmap. From there, you can access the GitHub repo and a static landing page summarizing the system.

To wrap up, this application was fully functional in December, but needed a major refactor to improve scalability and modularity. Each workflow is designed to be adaptable and plug-and-play, with clean separation between components and agents. The refactor is now underway and should only take a couple of days.

For updates, visit our Medium or Substack pages. The developer welcomes feedback and support from the quant community — through GitHub, article comments, Substack chat, or even by subscribing to the paid tier.

Thanks for watching, and see you soon for the next update.