Quantitative Researcher & ML Engineer

Building sophisticated models that transform market volatility into actionable insights. Proven track record of delivering alpha through advanced machine learning and rigorous quantitative research.

95%
Prediction Accuracy
$50M+
AUM Managed
15+
Production Models

Stefan Talpa

BS Computer Science, Georgia Institute of Technology, US

Concentrations: Modeling & Simulation, Theory

Scroll to explore

Core Expertise: Where Mathematics Meets Markets

Combining deep theoretical knowledge with practical implementation experience to solve complex financial challenges

Volatility Modeling

Expert in GARCH models, stochastic volatility, and implied volatility surfaces. Developed production systems achieving 95%+ accuracy in volatility forecasting.

  • GARCH/EGARCH
  • Stochastic Vol
  • Vol Surface Construction
  • Risk Metrics

Machine Learning for Finance

Building state-of-the-art ML pipelines for alpha generation, risk prediction, and market microstructure analysis using modern deep learning architectures.

  • Deep Learning
  • Feature Engineering
  • Ensemble Methods
  • Time Series

Algorithmic Trading Systems

End-to-end design and implementation of low-latency trading systems with robust risk management and real-time portfolio optimization.

  • Execution Algorithms
  • Signal Generation
  • Portfolio Optimization
  • Backtesting

Risk Analytics

Comprehensive risk framework development including VaR, CVaR, stress testing, and scenario analysis for multi-asset portfolios.

  • VaR/CVaR
  • Stress Testing
  • Attribution Analysis
  • Compliance

Quantitative Research

Data-driven research translating academic theory into profitable trading strategies with rigorous statistical validation.

  • Statistical Arbitrage
  • Factor Models
  • Market Microstructure
  • Research Pipeline

Production ML Infrastructure

Building scalable, maintainable ML systems from research to production with modern MLOps practices and monitoring.

  • MLOps
  • Cloud Infrastructure
  • Real-time Processing
  • Model Monitoring

Professional Experience: Delivering Measurable Results

Proven track record across leading financial institutions and innovative fintech startups

Senior Quantitative Researcher

Millennium Management

2023 - Present

Leading volatility modeling initiatives for equity derivatives desk managing $50M+ AUM. Developed proprietary ML models for volatility surface construction and options pricing.

  • Built LSTM-based volatility forecasting model achieving 95% directional accuracy
  • Reduced portfolio VaR by 23% through enhanced risk analytics framework
  • Implemented real-time options pricing system processing 100k+ quotes/second
  • Generated $2.3M additional alpha through improved execution algorithms
Python PyTorch C++ KDB+/Q AWS

Machine Learning Engineer

Citadel Securities

2021 - 2023

Developed production ML systems for market making and trade execution optimization. Focused on low-latency inference and real-time portfolio risk management.

  • Designed gradient boosting ensemble reducing prediction latency from 45ms to 8ms
  • Built feature engineering pipeline processing 10TB+ daily market data
  • Improved fill rates by 18% through optimal execution ML models
  • Created comprehensive monitoring dashboard reducing model debugging time by 60%
Python TensorFlow C++ Kubernetes Apache Kafka

Quantitative Analyst

Two Sigma Investments

2019 - 2021

Research and development of statistical arbitrage strategies across global equity markets. Built and maintained production trading systems with rigorous backtesting frameworks.

  • Developed cross-sectional momentum strategy with 2.1 Sharpe ratio in backtesting
  • Implemented factor model framework analyzing 5000+ securities daily
  • Reduced research-to-production cycle from 6 weeks to 2 weeks through automation
  • Created attribution system providing real-time P&L decomposition
Python R SQL Spark Docker

Research Intern - Machine Learning

Jane Street Capital

Summer 2018

Explored deep learning applications for market microstructure prediction. Prototyped novel architectures for order book dynamics modeling.

  • Built CNN-based model for limit order book prediction (78% accuracy)
  • Analyzed tick-level data across 200+ liquid securities
  • Presented research findings to trading desk and engineering teams
Python PyTorch OCaml Pandas

Let's Work Together

Legally authorized to reside and work in the European Union, with the ability to collaborate either as a full-time employee or as a self-employed contractor. Operates as a registered self-employed professional in Portugal and is based in Coimbra.

Available for consulting and full-time opportunities