What I Built
Built an end-to-end quantitative portfolio optimization platform for 1,000+ NSE-listed Indian equities. Given a basket of stocks, a historical look-back window, and an investment amount, CapiPort computes mathematically optimal capital allocation using two complementary frameworks β Efficient Frontier (4 objectives: max Sharpe, min volatility, efficient risk, efficient return) and Hierarchical Risk Parity. Deployed on Streamlit Cloud and HuggingFace Spaces.
What I Learned
HRP beats classical mean-variance when covariance estimates are noisy. Efficient Frontier is elegant but fragile β it inverts the covariance matrix, so estimation error blows up weights. HRP uses hierarchical clustering and recursive bisection to allocate capital without matrix inversion, producing more stable, diversified portfolios. Building both methods taught me the practical gap between textbook finance and real-world optimization.
Key Results
| Feature | Detail |
|---|---|
| Universe | 1,000+ NSE-listed equities (large, mid, small-cap) |
| Optimization methods | Efficient Frontier (4 objectives) + HRP |
| History | Flexible look-back from 1947 onward |
| Analytics | Annual/monthly returns, cumulative performance, per-stock contribution |
| Deployment | Streamlit Cloud + HuggingFace Spaces |
Project
π Live Demo - π€ HuggingFace | π GitHub Repository
Tech Stack: Streamlit, PyPortfolioOpt, yfinance, Plotly, MongoDB, GitHub Actions CI/CD | License: MIT
Citation
@online{prasanna_koppolu,
author = {Prasanna Koppolu, Bhanu},
title = {CapiPort {V2}},
url = {https://bhanuprasanna2001.github.io/projects/capiport.html},
langid = {en}
}