Krisoye Smith

Quantitative Researcher | Data Scientist | AI/ML Engineer

15+ years applying machine learning and AI techniques to quantitative finance, building predictive models across global equity markets, and developing high-performance tools for data science workflows.

Krisoye Smith - Quantitative Researcher

About Me

I'm a Quantitative Researcher with expertise in designing, prototyping, and deploying predictive models across global equity universes. My work spans quantitative finance, fraud detection, alpha/risk modeling, and building scalable ML/AI systems.

Most recently, as Head of Quantitative Research at Victory Capital (Sophus Capital), I maintained the Global EM alpha factor model. I've also led data science teams at Macquarie Asset Management and CO-OP Financial Services, driving innovation in predictive modeling and ML deployment.

I hold graduate degrees from NYU (MS Financial Engineering, MS Computer Science), University of Chicago Booth (MBA), and Iowa State University (MS Statistics).

Builder of production AI systems including RAG pipelines, vector search engines, and audio/document analysis servers deployed on GPU-accelerated infrastructure.

Technologies & Skills

Programming Languages

Python R SQL

Machine Learning & AI

PyTorch scikit-learn XGBoost Random Forest Elastic Net GAM

GenAI & RAG

LLMs RAG Pipelines Vector Databases Embeddings Claude API OpenAI Whisper MCP Prompt Engineering

Quantitative Finance

Factor Modeling Risk Management Portfolio Optimization Time Series FactSet Axioma

Cloud & Infrastructure

AWS Azure Databricks Linux

Data Engineering

pandas Arrow PySpark dplyr Airflow MLflow

Tools & Platforms

Git Docker LaTeX Jupyter

Featured Projects

Document Analysis MCP

MCP server for PDF processing powered by Claude API. Multi-page iterative extraction, document classification across 17+ content types, OCR support via Tesseract, and hash-based caching to prevent re-processing.

Claude API FastMCP pdfplumber Tesseract Python

Audio Analysis MCP

Production MCP server for audio analysis. Whisper transcription with word-level timestamps, pyannote.audio speaker diarization, prosody analysis, and sentiment detection. GPU-accelerated with low-VRAM mode for constrained hardware.

Whisper pyannote.audio FastMCP PyTorch CUDA

etlr

A comprehensive R package for streamlined ETL operations. Provides tools for data transformation, file management, cloud storage (S3) integration, and feature engineering. Built on dplyr, lubridate, and arrow for efficient data processing pipelines.

R AWS S3 Arrow ETL

mdlr

An R toolkit for streamlined model development, training, and forecasting at scale. Unified interface for base R and parsnip models (glmnet, mgcv, rstanarm), with support for rolling/expanding windows, walk-forward validation, and tidy exports to partitioned datasets.

R Machine Learning Parsnip Backtesting

S&P 500 Systematic Volatility Signal

Principal Global Investors: Predictive volatility trading signal for S&P 500 used in variable annuity hedging strategy. Delivered 21% better risk-adjusted returns (Sharpe Ratio) from 2012-2016 vs peers by dynamically allocating between equity and cash targeting 15% max annual vol.

Time Series Risk Management GARCH Portfolio Optimization

Get In Touch

I'm always interested in discussing quantitative research, data science opportunities, and collaborating on innovative projects. Feel free to reach out!