Daksh Kumar.
Raleigh, NC · Open to Quant / AI roles

Daksh KumarQuant Finance — Research & Trading

Building volatility models, backtesting engines, and ML systems — from options desks to production recommendation engines.

View ProjectsResumeLinkedInGitHub
Guaranteed MaxDD bound (LETF hedge)
15.0%
Net Sharpe (short-variance carry)
1.84
VIX replication corr.
0.9909
Delta-hedge variance reduction
26.1×

Selected work

Quant finance & ML projects

Backtesting engines, volatility models, and derivatives pricing — charts below render each project's actual published result files (equity curves, calibration fits, validation series), exported unmodified from the source repositories.

North Carolina State University · May 2026 – Present

Optimal Hedging of Leveraged ETF Positions (Convexity Protection)

A multi-tranche hedging engine on Tesla's 2× leveraged ETF (TSLL) — short the LETF, long ATM calls, short OTM puts (a collar) — harvesting structural volatility decay with maximum drawdown capped ex ante by a closed-form, model-free bound. The bound is fixed before the data run; the realized equity path never breached it.

View interactive breakdownFull report (PDF)
PythonOptionsLETFCollarNo-Arbitrage BoundsBloombergBacktesting
Floor engine (collar)
Naked short TSLL
Indexed to 100 · real backtest output
Guaranteed MaxDD bound (ex ante)
15.0%
Bound breaches (960 days)
0
CAGR / Sharpe (collar)
+15.69% / 0.90
Unhedged-short MaxDD
−81.1%
North Carolina State University · Jun 2026

Variance Risk Premium & Variance-Swap Replication (SPX)

Replicates a 30-day SPX variance swap from option strips (model-free implied variance, CBOE VIX methodology), validates the synthetic index against the live VIX, measures the variance risk premium with strictly forward-looking realized variance, and backtests a cost-aware short-variance carry.

View interactive breakdown
OptionsVariance SwapsVIXDatabentoStreamlitRisk Premia
Live CBOE VIX
Replicated 30-day implied vol
Vol points · real validation data
VIX correlation
0.9909
Mean abs error
0.568 vol pts
Net Sharpe / CAGR
1.84 / 18.0%
Mean VRP (83.7% days > 0)
+3.75 vol pts
Independent project · Jun 2026

Interest Rate Derivatives Pricing Engine (Swaptions)

An end-to-end fixed-income engine in pure Python/NumPy — no SciPy: every numerical routine (bootstrap root-find, cubic spline, monotone-convex interpolation, NSS fit, inverse-normal CDF, quasi-random sequences) implemented directly. Bootstraps Treasury zero curves from FRED, calibrates Hull-White, prices European and Bermudan swaptions by Monte Carlo, and runs hedging backtests and stress tests.

View interactive breakdown
NumPyHull-WhiteMonte CarloLongstaff-SchwartzYield CurvesFRED
Zero rate
Instantaneous forward
NSS fit
Curve date 2026-06-15 · real engine output
Hull-White fit RMSE
2.75 bp
MC vs analytic error
0.022 bp
Bermudan exercise premium
80.4 bp
Hedge variance reduction
26.1×
North Carolina State University · Jan 2026 – May 2026

Calibration and Hedging of Local, Stochastic, and Hybrid Volatility Models

A unified calibration and hedging framework for Local (Dupire), Stochastic (Heston), and Stochastic-Local Volatility (SLV) models across 5,000+ SPX option contracts.

View interactive breakdown
DupireHestonSVIPDEDelta Hedging
Market IV
SVI model fit
Implied vol % · simulated
Surface RMSE
2.47%
BS Inversion Convergence
94%
Deep-OTM Error Reduction
~15%
P&L Leakage Reduction
12%

Career

Work experience

Founding engineering roles spanning ML-driven recommendation systems and real-estate data science.

Founding AI Engineer · CBCatalyst

Mar 2025 – Jan 2026

Delhi, India

  • Built and fine-tuned a multi-modal recommendation engine using CLIP, BART, and GPT-2 with custom ranking attributes (skew factor, rank decay), boosting Top-10 recall and CTR by 35% and 50% respectively.
  • Optimized high-dimensional CLIP embedding spaces using contrastive objectives, hard-negative sampling, and similarity calibration — a 37% improvement in Top-K ranking under noisy, near-duplicate data.
  • Designed a reasoning-based RAG system on Milvus (HNSW) for candidate retrieval, cross-encoder re-ranking, and guard-railed LLM inference for live AI search.

Founding Data Science Engineer · PropReturns (YC S21)

Jan 2023 – Mar 2024

Mumbai, India

  • Built a real-time Automated Valuation Model (AVM) with XGBoost across 1M+ nationwide transactions, achieving 73.8% PPE10 and R² of 0.79 by fusing internal data with public records.
  • Developed an automated legal-risk parsing engine (Marker OCR + Gemini API) to extract title disputes and encumbrances from unstructured court verdicts, linking risk profiles to the core property database.

Toolbox

Skills

Languages

PythonPandasNumPySciPyscikit-learnPyTorchTensorFlowCC++Java

Quant / Finance

Options & Derivative PricingStochastic CalculusLocal/Stochastic Vol (Dupire, Heston, SLV)Variance SwapsBacktestingBloomberg Terminal & API

Databases & Cloud

PostgreSQLMySQLAWS (Lambda, EC2, ECR)GCP (BigQuery)DockerMLflow

Protocols & Systems

LinuxBashGitREST APIWebSocketWebhook

Background

Education

North Carolina State University (NCSU)

Master of Financial Mathematics

Dec 2027
Raleigh, NCGPA 3.66 / 4.0
Linear Matrix & TransformationLinear AlgebraPartial Differential EquationsNumerical MethodsFinancial Market OperationsOptions and Derivative PricingProbability and Stochastic for Finance
View NCSU faculty/student profile

Jawaharlal Nehru University (JNU)

B.Tech, Electronics & Communication Engineering + M.Sc. Management (Dual Degree)

June 2023
New Delhi, IndiaGPA 3.64 / 4.0

Get in touch

Let's talk quant, ML, or your next hire.

Currently based in Raleigh, NC and open to quant research, quant dev, and applied AI roles. The fastest way to reach me is email.

dakshdhull@gmail.comLinkedInGitHubResume