Looking for a job

Alexander Booth

@alexanderbooth

Message

I build production agentic AI workflows and code-generation capabilities for developers.

United States

Message

At GitHub, I build and productionize Copilot code-generation capabilities using PyTorch, OpenAI, prompt optimization, contextual ranking, and telemetry. I develop repository-aware RAG, semantic retrieval, and coding-agent workflows that help developers delegate bug fixes, refactoring, and feature implementation.

I establish evaluation, safety, and guardrails through trajectory scoring, task-success metrics, test and linter feedback, policy enforcement, and human-in-the-loop PR review. I also partner with product, design, security, and business stakeholders to turn complex LLM tradeoffs into measurable product decisions.

Previously, I built forecasting, inventory optimization, customer segmentation, pricing, cyber-risk scoring, and claims models for Specialized Bicycle Components and Corvus Insurance using Python, SQL, XGBoost, scikit-learn, and statistical modeling.

Experience

Work history, roles, and key accomplishments

Current

Staff Machine Learning Engineer

Current

GitHub

Aug 2021 - Present (5 years 1 month)

Built and productionized GitHub Copilot code-generation capabilities using PyTorch, OpenAI, prompt optimization, contextual ranking, and telemetry. Developed repository-aware RAG and semantic retrieval, and engineered Copilot coding agent workflows with LLM reasoning and tool calling.

PyTorch OpenAI prompt optimization Contextual Ranking Telemetry RAG Semantic Retrieval Vector Search Reranking Context Selection LLM Evaluation Intelligent Routing A B Testing Lang Chain LangGraph Agent orchestration AI Observability GitHub Actions Testing Automation

Senior Data Scientist

Specialized Bicycle Components

Jun 2020 - Jul 2021 (1 year 1 month)

Built SKU demand forecasting with XGBoost and Prophet, improving inventory allocation and reducing lost-sales risk. Developed inventory optimization models and customer propensity models to improve conversion efficiency.

XGBoost Prophet Demand Forecasting Inventory Optimization Probabilistic Forecasting SQL Python scikit learn Gradient Boosting Customer Segmentation regression Causal Inference A B Testing

Machine Learning Engineer

Corvus Insurance

Sep 2019 - May 2020 (8 months)

Built cyber risk scoring models using XGBoost and scikit-learn to predict security incidents, improving underwriting risk selection. Developed claims probability models from security telemetry and policy data.

XGBoost scikit learn Risk Scoring Feature Engineering Statistical Modeling Security Telemetry Policy Data Underwriting