The latest articles on DEV Community by Abhinav (@abhinavsharma11pix). https://dev.to/abhinavsharma11pix https://media2.dev.to/dynamic/image/width=90,height=90,fit=cover,gravity=auto,format=auto/https:%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Fuser%2Fprofile_image%2F4026337%2F3c111bff-3668-4da2-a495-429bc6222605.png https://dev.to/abhinavsharma11pix en Abhinav Sun, 12 Jul 2026 18:42:06 +0000 https://dev.to/abhinavsharma11pix/ai-portfolio-analyzer-6il https://dev.to/abhinavsharma11pix/ai-portfolio-analyzer-6il <h1> Building an AI Portfolio Analyzer with Python, FastAPI, React &amp; Machine Learning </h1> <p>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.</p> <p>The result is <strong>AI Portfolio Analyzer</strong>, a web application that helps investors analyze portfolios, forecast prices using machine learning, understand portfolio risk, and estimate capital gains taxes.</p> <p>🔗 <strong>Live Demo:</strong> <a href="https://portfolio-analyzer-sigma-amber.vercel.app/" rel="noopener noreferrer">https://portfolio-analyzer-sigma-amber.vercel.app/</a></p> <p>💻 <strong>GitHub:</strong> <a href="https://github.com/abhinavsharma11pix/portfolio-analyzer" rel="noopener noreferrer">https://github.com/abhinavsharma11pix/portfolio-analyzer</a></p> <h1> Why I Built It </h1> <p>Most portfolio tracking applications focus on showing holdings and returns. I wanted to build something that could answer questions like:</p> <ul> <li>How risky is my portfolio?</li> <li>What does the historical performance tell me?</li> <li>What could prices look like over the next month?</li> <li>How much tax would I owe if I sold today?</li> <li>Can AI help summarize portfolio performance?</li> </ul> <p>This project became an opportunity to combine quantitative finance, machine learning, and full-stack engineering into a single application.</p> <h1> Features </h1> <h2> Live Market Data </h2> <p>The application supports both NSE and NYSE stocks with real-time market updates through WebSockets.</p> <p>This enables live portfolio valuation without constantly refreshing the page.</p> <h2> Machine Learning Forecasting </h2> <p>One of the most interesting parts of the project was building the forecasting pipeline.</p> <p>Instead of relying on a single model, I combined multiple approaches:</p> <ul> <li>Exponential Smoothing (ETS)</li> <li>Random Forest Regressor</li> <li>LightGBM</li> </ul> <p>Each model captures different characteristics of market behavior.</p> <p>The application generates a 30-day forecast and presents it alongside historical prices for comparison.</p> <h2> AI Portfolio Advisor </h2> <p>The application uses <strong>Llama 3</strong> to generate portfolio insights.</p> <p>Rather than simply explaining metrics, it analyzes portfolio composition and produces natural language summaries that are easier for investors to understand.</p> <p>Examples include:</p> <ul> <li>concentration risk</li> <li>diversification opportunities</li> <li>sector exposure</li> <li>portfolio observations</li> </ul> <h2> Portfolio Risk Analytics </h2> <p>The platform calculates several commonly used portfolio metrics including:</p> <ul> <li>Sharpe Ratio</li> <li>Value at Risk (VaR)</li> <li>Sortino Ratio</li> <li>Beta</li> <li>Maximum Drawdown</li> </ul> <p>These metrics provide additional context beyond simple profit and loss.</p> <h2> Capital Gains Tax Engine </h2> <p>Another feature I wanted to include was tax estimation.</p> <p>The application supports:</p> <ul> <li>FIFO lot matching</li> <li>Short-term and long-term capital gains</li> <li>Tax estimation</li> <li>Harvesting insights</li> </ul> <p>This allows investors to understand the tax implications of selling holdings.</p> <h1> Tech Stack </h1> <h3> Backend </h3> <ul> <li>Python</li> <li>FastAPI</li> <li>SQLite</li> <li>WebSockets</li> </ul> <h3> Machine Learning </h3> <ul> <li>Scikit-Learn</li> <li>LightGBM</li> <li>Pandas</li> <li>NumPy</li> </ul> <h3> Frontend </h3> <ul> <li>React</li> <li>TypeScript</li> <li>Vite</li> </ul> <h3> Deployment </h3> <ul> <li>Vercel</li> <li>Render</li> </ul> <h1> Challenges </h1> <p>Every project comes with tradeoffs.</p> <p>Some of the interesting challenges included:</p> <ul> <li>Designing APIs that could efficiently serve portfolio data.</li> <li>Keeping market data responsive.</li> <li>Combining statistical and machine learning models into a single forecasting workflow.</li> <li>Balancing prediction quality with inference time.</li> <li>Building an intuitive dashboard that exposes complex financial metrics without overwhelming users.</li> </ul> <p>Each iteration improved both the user experience and the architecture.</p> <h1> Lessons Learned </h1> <p>Building this project strengthened my understanding of:</p> <ul> <li>designing production-ready APIs</li> <li>machine learning workflows</li> <li>frontend and backend integration</li> <li>financial analytics</li> <li>performance optimization</li> <li>deploying full-stack applications</li> </ul> <p>More importantly, it reminded me how much you learn by building end-to-end products rather than isolated models.</p> <h1> What's Next? </h1> <p>Some ideas I'm currently exploring include:</p> <ul> <li>Portfolio optimization algorithms</li> <li>Additional forecasting models</li> <li>User authentication and cloud database support</li> <li>Backtesting strategies</li> <li>More interactive visualizations</li> </ul> <p>If you have suggestions or feedback, I'd love to hear them.</p> <p>⭐ GitHub:<br> <a href="https://github.com/abhinavsharma11pix/portfolio-analyzer" rel="noopener noreferrer">https://github.com/abhinavsharma11pix/portfolio-analyzer</a></p> <p>🚀 Live Demo:<br> <a href="https://portfolio-analyzer-sigma-amber.vercel.app/" rel="noopener noreferrer">https://portfolio-analyzer-sigma-amber.vercel.app/</a></p> fintech machinelearning python showdev