I'm a Computer Science and Applied Mathematics double major at Vanderbilt University (4.0 GPA, Minor in Data Science) graduating May 2026. My focus is on quantitative research and market microstructure — I'm driven by the challenge of extracting signal from noisy financial data using rigorous mathematical modeling and high-performance computing. I have hands-on experience building ultra-low latency trading infrastructure at Morgan Stanley's Speedway Team and conducting original research at the intersection of statistical learning and financial markets.

📍 Location: Basking Ridge, NJ 🎓 Education: Vanderbilt University (B.A. in CS & Applied Mathematics, Minor in Data Science — May 2026) 📬 Email: cameronscarp@gmail.com 🔗 LinkedIn: linkedin.com/in/cameron-scarpati

Awards: CRA Research Honorable Mention | Provost's Faculty Grant for Immersion Vanderbilt ($2000) | Calculus Top-Student Award | Dean's List (All Semesters)

Relevant Coursework: Data Structures | Algorithms | Machine Learning | Financial Mathematics | Linear Optimization | Probability & Statistics | Differential Equations | Linear Algebra | Operating Systems | Computer Architecture

A market microstructure analytics platform that detects hidden trading regimes from cryptocurrency order book data using Gaussian Hidden Markov Models. Processes Level 2 order book snapshots and extracts 30+ microstructure features (OFI, VPIN, Kyle's λ, book imbalance, realized volatility at multiple scales) to identify three distinct market states — Quiet, Trending, and Toxic. Key finding: VPIN spikes 30–120 seconds before toxic regime transitions, serving as a leading indicator of adverse selection. A regime-conditional strategy achieves a 1.8–2.5 annualized Sharpe ratio. Built with Python 3.11+ and a C++17/pybind11 LOB engine processing 1M+ updates/sec, with an interactive Plotly Dash dashboard.

An arbitrage-free implied volatility surface construction engine for live SPY options. Fetches real-time options chains via yfinance, extracts implied volatility using Newton-Raphson root-finding (with Brent's method fallback, convergence to |Δσ| < 10⁻¹⁰), and fits the SVI parameterization per expiry slice via multi-start L-BFGS-B optimization. Enforces Durrleman butterfly and calendar-spread monotonicity no-arbitrage constraints, computes Dupire local volatility and full Black-Scholes Greeks (Δ, Γ, ν, Θ). Achieves <0.5 vol-point RMSE across all slices while identifying 15–20 statistically significant mispricings per snapshot. Interactive Streamlit dashboard with 3D surface plots, delta-space smile views, Greeks heatmaps, and arbitrage diagnostics. Built with Python, NumPy, SciPy, and Plotly; tested with 130 pytest tests and CI via GitHub Actions.

A C++17 AI agent that plays the game Buckshot Roulette using Expectiminimax search with alpha-beta pruning. Evaluates thousands of game states per move within a 7-second budget via iterative deepening (depths 5–20). The game tree models max nodes (bot), min nodes (opponent), and chance nodes (shell uncertainty), with leaf scoring driven by weighted heuristics: health differential (600 pts), shell knowledge (300 pts), and item valuations (15–40 pts). Demonstrates adversarial search under uncertainty — a core algorithmic pattern in quantitative decision-making.

An interactive C++/OpenGL visualization of the Collatz conjecture — one of mathematics' most famous unsolved problems. Supports bulk sequence generation and targeted number selection with animated gradient rendering, logarithmic/linear axis toggling, and real-time statistical overlays (step counts, peak values, averages). Performance-optimized through memoization for constant-time lookups on previously computed sequences.

Morgan Stanley · Equity Algorithms June 2025 – August 2025 · New York, NY

• Developed for the Speedway Team, an ultra-low latency solution connecting institutional clients to stock exchanges supporting up to 25,000 orders per second for high-frequency trading and market making
• Revamped stress-testing framework across Client Connectivity Services increasing throughput up to ~45%
• Researched C++ to push the limits of sending TCP messages at speeds of up to ~3–4 million per second

LendOS · NestJS | React | DAML | Agile & Scrum June 2024 – August 2024 · Remote

• Engineered components of a groundbreaking commercial lending platform optimizing its scalability and robustness
• Partnered closely with the product team to ensure the platform's user interface and functionality aligned seamlessly with strategic business goals

Vanderbilt University · Java August 2024 – December 2024 · Nashville, TN

• Conducted office hours to mentor students, providing personalized guidance on course materials and assignments

Morgan Stanley · Technology February 2024 · New York, NY

• Engaged in workshops to explore operational strategies and technology-driven solutions to financial challenges

• Deep Learning Side-Channel Analysis — Primary Author | ITiCSE 2026 (Manuscript in Preparation)
• Alternative Risk Measures and Stock Selection — Editorial Contributor | Under Review (2025)
  • Delivered editorial guidance and LaTeX formatting for research on entropy and Hurst exponent as volatility measures

Club Tennis Team | Conversational Italian | Rock Climbing | Chess | Sudoku | Table Tennis | Pool