Hornet FX desk
A USD/MXN worksheet adapted from my trading-contest project. Edit trade rows, reprice the book, and inspect combined risk.
Open the worksheetProjects
Financial tools, quantitative research, and experiments you can try.
Explore the vibe coded experiments for small visual studies and the next demos.
Try a project
Browser language model
A 30-million-parameter model running locally in your browser. An inference experiment, with a very small context and imperfect answers.
Try the modelSelected work
Implied Willow Tree for Derivatives
A study of risk-neutral densities from 10Y Treasury futures options, with martingale and no-arbitrage constraints for a new objective function.
Python, numerical optimization, Treasury futures options
Project details
- Extends the willow-tree idea from equity options toward a liquid interest-rate derivatives market.
- Uses capital-markets information and discretized risk-neutral nodes to study path-dependent pricing.
High Frequency Trading Competition
A second-place market-making strategy that used VWAP to predict trends and adapt order size.
Python, VWAP, SHIFT, FIX protocol, GitHub
Project details
- Led the team, delegated backtesting and research, and kept the group aligned through meeting notes.
- Built backtesting and algorithm components and acted as the point of contact for SHIFT and FIX issues.
Hornet Trading Contest
A risk-engineering trading contest covering zero-coupon bonds, forwards, calls, and puts on USDMXN.
VBA, Bloomberg, bond and FX derivatives
Project details
- Wrote VBA to calculate Delta, Gamma, USD and MXN Rho, and Vega for traded instruments.
- Used current-events information from Bloomberg to anticipate a falling dollar and adjust delta exposure.
NBA Defensive Impact Analysis
A data pipeline that cleaned gigabytes of API data, cached rate-limited requests, and moved analysis into SQL for Power BI.
Python, APIs, Pandas, SQL, MSSQL, Azure SQL, Power BI
Project details
- Removed unnecessary columns and changed data types to improve memory use.
- Added backoff and caching for repeated API collection, then synchronized SQL tables to Azure SQL.
NBA Reddit AI Chatbot
A chatbot over daily NBA subreddit posts and comments, tuned to understand community slang and nested discussion structure.
Python, LangChain, retrieval-augmented generation
Project details
- Built a LangChain retrieval pipeline over the collected post and comment corpus.
- Used prompt engineering to make answers preserve the source community's terminology and context.