Python and SQL analysis of 23,246 NFL records, with ridge regression evaluated against three baselines on 7,847 player-week forecasts. The research covers 2022–2026; 2026 is a partial snapshot through Week 4. The interactive lineup lab uses a selected-player 2024 replay with exact seven-slot optimization. Reference data and outputs were stored in AWS S3, with a replicated dataset processed in Databricks using PySpark and Delta tables.

Python validation, feature engineering, logistic regression, and regenerated dashboard outputs. Evaluation uses a separate holdout; the portfolio view identifies training and held-out rows. Includes AWS S3 and Databricks integration code, not a claimed production deployment.

Explore sourced financial research, sensitivity analysis, capital structure, and modeled equity returns. A hypothetical acquisition case using public data; projected results are not realized investment returns.

This project investigates whether structured, engineering-based techniques can outperform passive benchmarks in algorithmic trading.

Research Objectives:

• Apply mathematical modeling and calculus to extract actionable structure from market data
• Develop a signal generation system using slope and acceleration of price curves
• Optimize entry/exit thresholds for dynamic market regimes
• Benchmark performance against passive strategies like SPY and QQQ

• Uses Savitzky–Golay filters to smooth price curves while preserving turning points
• Computes first (slope) and second (curvature) derivatives to detect inflection trends
• Applies threshold logic to trigger trades based on derivative magnitude and direction
• Performs backtests with Sharpe ratio, drawdown, ROI, win rate, and trade count
• Benchmarked against SPY buy-and-hold to assess relative edge

• Real-time ticker selection and signal overlay via Streamlit
• Moving slope and acceleration visualizations
• Heatmap-based parameter optimization with adjustable entry/exit zones
• Walk-forward validation and fold-specific Sharpe analysis
• Trade log download and performance history export

Built as part of the Data-Driven Decisions course, this project evaluates the variability and potential of wind energy using Monte Carlo simulation.

Study Goals:

• Model uncertainty in wind turbine output across different regions
• Compare U.S. renewable capacity to Germany, Denmark, and the UK
• Project renewable potential under optimistic and policy-driven scenarios
• Visualize distributions, outliers, and projected capacity curves

• Generated thousands of wind output scenarios using Monte Carlo simulation in Python
• Used numpy.random.normal to represent wind performance variability
• Compared real-world vs. theoretical performance using histograms and overlaid mean curves
• Incorporated region-specific assumptions about capacity, wind strength, and policy ambition

• U.S. performance lags behind Europe due to underutilization and slower growth rate
• Monte Carlo results quantify uncertainty and help guide future energy investment decisions
• Supports policy discussions on renewables by grounding them in statistical simulation

This project simulates the backbone of a high-frequency or signal-based trading system, using Rust for performance and system control.

Core Capabilities:

• Models how a real-time engine ingests and processes synthetic market data
• Performs intensive floating-point computation using array simulation logic
• Prints dynamic output such as simulated Sharpe ratios and momentum scores
• Exposes a clean modular architecture via main.rs, gpu.rs, and mod.rs

• Simulates compute-heavy tasks typical in quantitative research or order execution
• Structured to be extensible — future GPU acceleration and real-time integration possible
• Demonstrates strong command of systems-level performance techniques in Rust