JCJoshua Colmenar

Quantitative finance research

Derivatives, risk, and portfolio construction.

I build research implementations to study how models respond to conventions, constraints, and imperfect data. Current work covers SOFR curves and swaps, constrained portfolio construction and attribution, and option pricing.

MSc Quantitative FinancePython + C++New Jersey
View research GitHub

Selected work

Research projects

Each project page separates the method, the implementation, the evidence, and the limits of the available data.

01 / Interest-rate derivatives

SOFR Curve Construction, Swap Risk & Value-Change Explain

Sequentially bootstraps a 1W-30Y SOFR OIS curve, values fixed-versus-daily-compounded-SOFR swaps, and rebuilds the curve for parallel DV01, key-rate DV01, and full-revaluation shocks.

Data boundary: OIS quotes and trades are synthetic. New York Fed SOFR and SOFR Index observations are real; Treasury yields are context only.

  • Log-DF bootstrap
  • ACT/360 compounding
  • Swap valuation
  • DV01
  • Value-change explain
1W-30Ycurve pillars
4curve scenarios
0.0explain identity difference
Project pageSourceSample workbook
Bootstrapped synthetic SOFR zero and forward curves with discount factors
Sequential bootstrap output from the published 2 July 2026 sample.

02 / Portfolio construction

Constrained Portfolio Construction & Factor Attribution

Compares seven long-only allocation methods under lagged walk-forward estimation, position limits, tracking error, turnover, transaction costs, factor-exposure controls, and exact-cardinality constraints.

Data boundary: Website figures use the deterministic synthetic demonstration. The retained CRSP-derived price panel is evaluated locally and is not published.

  • Ledoit-Wolf
  • Mean-variance
  • Historical CVaR
  • CVXPY
  • HiGHS MILP
  • FF5 + momentum
7long-only methods
72monthly decisions
8exact holdings in CVaR MILP
Project pageSourcePublic summary
Synthetic portfolio methods compared by annualized return, volatility, and turnover
Public deterministic demonstration. The chart illustrates method behavior, not investment performance.

03 / Option pricing

Volatility Surface Calibration & Option Pricing

Moves from saved option chains through carry, contract routing, implied-volatility inversion, raw SVI calibration, static-arbitrage repair, and European or American pricing with numerical cross-checks.

Data boundary: The public release excludes licensed quote chains. It includes aggregate evidence, figures, source registries, and synthetic workflows.

  • Put-call parity
  • Brent roots
  • Raw SVI
  • PAVA + SLSQP
  • CRR / LR
  • CN-PSOR
48asset-date contexts
2,400valid closing prices
7strict quote-quality failures retained
Project pageSourceTechnical note
Calibrated NDX implied-volatility surface
NDX calibrated surface in forward log-moneyness and maturity; interpolation is in total variance.

Across projects

Research practice

The common focus is not a single model. It is the chain from assumptions and data to numerical results and diagnostics.

ProblemEstimation or modelImplementationCheck
SOFR curve and swapsSequential OIS bootstrap; compounded overnight legBrent roots; log-DF interpolation; full revaluationQuote repricing, index reconciliation, explain identity
Portfolio constructionShrunk means; Ledoit-Wolf covariance; historical CVaRCVXPY continuous problems; SciPy/HiGHS MILPLag audit, constraint residuals, block bootstrap
Option pricingParity, SVI, arbitrage repair, early exercisePython research path; C++ tree and PDE libraryPaired contexts, model ladders, bound and quality gates

Data

Provenance before interpretation

Data status and publication boundaries are stated separately: real and synthetic inputs, licensed data excluded from public assets, and retained data kept local. None of the current workflows requires a live WRDS connection.

Timing

Decisions use available information

Portfolio weights are formed from lagged windows. Curve and option results are tied to dated snapshots and explicit conventions.

Evidence

Failures remain visible

Fallbacks, rejected formulations, residual checks, and known limitations remain part of the project record rather than being removed from the presentation.

Profile

Joshua Colmenar

Quantitative finance researcher focused on derivatives valuation, portfolio construction, risk measurement, and numerical implementation.

EducationMSc Quantitative Finance, Rutgers Business School, 2026
UndergraduateBS Finance and Economics, Kean University, 2024
ExperienceAdministrative Analyst, New Jersey Department of Health, 2025
ToolsPython, C++, pandas, NumPy, SciPy, CVXPY, PyTorch, SQL, Excel
LocationUnion, New Jersey