Over the last few months, I wanted to build a project that combined everything I enjoy working on—data engineering, machine learning, backend APIs, and modern frontend development.
The result is AI Portfolio Analyzer, a web application that helps investors analyze portfolios, forecast prices using machine learning, understand portfolio risk, and estimate capital gains taxes.
🔗 **Live Demo:** [https://portfolio-analyzer-sigma-amber.vercel.app/](https://portfolio-analyzer-sigma-amber.vercel.app/)
💻 **GitHub:** [https://github.com/abhinavsharma11pix/portfolio-analyzer](https://github.com/abhinavsharma11pix/portfolio-analyzer)
Most portfolio tracking applications focus on showing holdings and returns. I wanted to build something that could answer questions like:
This project became an opportunity to combine quantitative finance, machine learning, and full-stack engineering into a single application.
The application supports both NSE and NYSE stocks with real-time market updates through WebSockets.
This enables live portfolio valuation without constantly refreshing the page.
One of the most interesting parts of the project was building the forecasting pipeline.
Instead of relying on a single model, I combined multiple approaches:
Each model captures different characteristics of market behavior.
The application generates a 30-day forecast and presents it alongside historical prices for comparison.
The application uses Llama 3 to generate portfolio insights.
Rather than simply explaining metrics, it analyzes portfolio composition and produces natural language summaries that are easier for investors to understand.
Examples include:
The platform calculates several commonly used portfolio metrics including:
These metrics provide additional context beyond simple profit and loss.
Another feature I wanted to include was tax estimation.
The application supports:
This allows investors to understand the tax implications of selling holdings.
Every project comes with tradeoffs.
Some of the interesting challenges included:
Each iteration improved both the user experience and the architecture.
Building this project strengthened my understanding of:
More importantly, it reminded me how much you learn by building end-to-end products rather than isolated models.
Some ideas I'm currently exploring include:
If you have suggestions or feedback, I'd love to hear them. ⭐ GitHub:
https://github.com/abhinavsharma11pix/portfolio-analyzer 🚀 Live Demo:
https://portfolio-analyzer-sigma-amber.vercel.app/