At Fractal Labs, we set out to solve a key challenge for individual day traders: the lack of accessible, high-quality tools to predict stock trends. Our solution, Your Personal Quant (working name), is an AI-powered investment app designed to forecast stock movements by combining machine learning and sentiment analysis—an advantage typically reserved for large financial institutions. We wanted to bring this capability to the masses.
This app’s story began with one of our summer interns, Matthew Beck, who developed a promising proof of concept as part of a portfolio project. Inspired by his vision, we expanded his work, drawing on our product and engineering strengths to create a robust, feature-rich tool aimed at empowering traders with insights they wouldn’t otherwise have. This case study shares our journey in building Your Personal Quant—from its origins and technical innovations to the challenges and lessons learned along the way.
The concept behind Your Personal Quant was simple: enable amateur day traders to make informed trading decisions with the same predictive power that institutional investors possess. Day traders, who often buy and sell stocks within short windows, usually rely on specialized tools or teams to gain insights. We aimed to break this barrier by blending machine learning models with natural language processing to provide real-time, accessible insights to the individual trader.
The app forecasts stock movements over short periods, like a day or a few days, by analyzing historical data and supplementing it with sentiment analysis from news sources. This approach offers users a more holistic view of potential stock movements, enabling them to make data-driven decisions.
When intern Matthew Beck introduced his working proof of concept, it offered basic functionality, analyzing historical stock data to predict price movements. Recognizing its potential, we set out to expand his foundation into a full-featured product. Your Personal Quant became our first project to incorporate a custom machine learning model built entirely in-house, a departure from our previous reliance on third-party AI solutions (we incorporated those third-party solutions to help simplify the qualitative data too!).
We used TensorFlow to process and model stock data, crafting an ML model tailored to predict stock trends from both quantitative and qualitative data. This work not only helped us refine our approach to AI model development but also prepared us for future projects requiring complex data integration.
The defining feature of Your Personal Quant is its ability to integrate both quantitative stock data and qualitative sentiment data for a comprehensive view. Its standout features include:
This blend of analysis ensures that Your Personal Quant delivers actionable insights, helping traders make fast, informed decisions.
Early on, we encountered the challenge of focusing on core functionality over secondary features. Many teams fall into the trap of allocating too much time to standard app functions like user sign-ups and billing before the main value is even validated. As Matt Lim, our Product Manager, emphasized, we avoided this by directing resources toward refining the prediction model and validating its accuracy, putting off infrastructure elements until the app’s core value was fully realized.
A critical milestone for Your Personal Quant was the successful creation of a custom machine learning model and its integration with sentiment analysis specific to financial news. Supported by AWS and built with TensorFlow, this pipeline allowed us to establish a scalable foundation for future projects that may require specialized AI models.
We also implemented fine-tuning for the sentiment component, adapting a large language model (LLM) to analyze financial news effectively. Fine-tuning enabled the model to capture and present financial nuances with greater accuracy, making the predictions more relevant and reducing costs associated with broader AI techniques. This fine-tuning expertise enhances our ability to tailor AI to specific applications, such as finance or other data-driven fields.
We are still in development but will launch a beta version soon. Currently, our focus is on refining model accuracy and gathering user feedback to enhance its capabilities. We’re initially testing with a limited number of stocks, planning to scale as we validate the core functionality. Once the core is solid, we’ll add administrative features, such as user accounts and billing systems, allowing us to expand infrastructure based on proven value.
Beyond this project, our advancements with Your Personal Quant open doors to new AI applications in fields like healthcare and legal services. The technical foundation we’ve built can be applied across sectors, creating possibilities for future AI-driven projects at Fractal Labs.
Your Personal Quant represents a leap forward for Fractal Labs in making powerful AI insights available to everyday users. By combining stock data with real-time sentiment analysis, we’re delivering predictive capabilities that empower day traders at the level of institutional investors. As we near the app’s launch, we’re excited to explore how this technology might further transform industries through accessible, AI-driven tools.
%20(1).png)
.png)
