code
hello! here are some of the projects I’ve worked on.
0.1 Macro-forecaster (working draft)
I created this web-app as a companion to a working paper. Overlays the IMF’s 5-year forecast path against Naive, ARIMA, ETS, Prophet, and Chronos (Amazon’s T5-based foundation model) forecasts, plus four ensemble combination methods including an adaptive inverse-MSE weighted combination.
0.2 Zetetic-LLM powered investment app
A Streamlit application that converts your investment thesis into executable strategies using multiple AI models, simulates those strategies with Yahoo Finance data, and provides comprehensive performance analysis with backtesting support and estimated AI cost transparency.
0.3 Discounted cash flow (DCF) analysis tool
I created this web-app to automate basic Monte Carlo DCF analysis for publicly traded companies (with data available via Yahoo Finance). Enter a ticker and it pulls live financials from Yahoo Finance, estimates WACC via CAPM, projects free cash flows, and calculates an intrinsic value per share.
It runs Monte Carlo simulations to quantify uncertainty, offering a sensitivity table color-coded against the market price, and backtesting the model against historical data to see if past signals were directionally correct. The risk-free rate is auto-fetched from FRED. Built with Python, Streamlit, and Plotly.
0.4 Implementation of Lilien Index (1982) in R and Python (working draft)
This Github repository provides clean, reference implementations of the Lilien Index, generally used for measuring sectoral employment dispersion and structural change. Implementations are provided in R and Python (forthcoming), with consistent definitions, shared examples, and unit tests.
0.5 U.S. Federal Expenditure tracker (Brookings)
For some interesting work I did with some colleagues (s/o Noadia + Lauren), check out this R Shiny app on U.S. Federal Outlays (spending)
0.7 U.S. Tariff & Trade tracker (Brookings)
During the height of the tariff uncertainty in Q1/Q2 2025, I developed a tariff tracker with my wonderful colleagues. Check it out!
0.8 Global private default risk analysis (working draft)
I created this web-app to perform predictive analytics that estimates private default risk at the country-level using World Bank GEMS default-rate data as the target variable and IMF Financial Development Index indicators, along with macroeconomic fundamentals, as the feature set. Technical features:
- Automated, versioned data ingestion from World Bank and IMF FDI APIs
- Modular components between data, models, and UI, with prediction logic and feature engineering implemented in R.
- Bootstrap-ensemble ML pipeline for distributional forecasting.
- Built-in model interpretability using SHAP values
- Reduced runtime latency through offline model training with hashed predictor sets and registry
- Automated reporting via R markdown, generating consistent “presentation-ready” briefs.
- Deployment practices using Git/GitHub for version control and ShinyApps.io/containerized hosting for reproducible delivery.
0.6 Social spending mini-simulation
I created this simple mini-simulation to visualize how redistribution through social spending works. It’s quite “naive” in the sense that it models the income distribution of an entire country as ten shares. However, it helped me understand how redistribution directly impacts inequality metrics like the Gini coefficient.