Nathan Stefanik

Experience

JetBlue Airways Jun 2023 – Present

Data Scientist · New York, NY

Statistical modeling for airline revenue management: interval-valued demand forecasts for pricing review, with the evaluation and validation layer around them.

• Built and still maintain the forecasting system for revenue-management pricing review — a staged Snowflake → Polars → R → Databricks → MLflow path that retrains, evaluates, and packages candidates as a system, not a notebook.
• Rebuilt the Python/R/Databricks boundary after it depended on local files and implicit run state: R keeps fit and score I/O; Python owns run lifecycle, artifacts, and registry, so models package as Databricks artifacts.
• Designed experiment comparison against a failure mode coverage hides: a wide prediction interval looks accurate. Width and tail behavior are first-class; Tableau is the review surface, so a single green coverage number is not quality.
• Made model identity load-bearing through scoring, post-processing, evaluation, and production-candidate tables, including data-preserving Snowflake migrations, so comparisons are between models and a reported figure traces to a specific artifact.
• Designed the validation layer between forecast and pricing action: a short, decision-oriented review that surfaces unsafe segments without a full-network read before a candidate is treated as production-ready.
• Caught a silent evaluation failure where related model variants shared SQL and mixed observed series — a candidate could pass against the wrong data — then made parameters task-owned so each variant validates against its own observations.

Invesco Ltd Jun 2022 – Aug 2022

Info Risk and Controls Intern · Atlanta, GA

• Analysis for the Global Risk and Controls team, including an incident-ticket classification model (Doc2Vec + GMM) at 70% accuracy.
• Processed large datasets in Python from Microsoft SQL Server (pyodbc).
• Built PowerApps for internal risk management and control validation.

Georgia Tech School of Mathematics Fall 2021

Grader, Linear Algebra II · Atlanta, GA

• Graded problem sets for over 70 students and wrote answer keys and rubrics.

Education

Georgia Institute of Technology Dec 2022

B.S., Mathematics and Computer Science · Atlanta, GA · GPA 3.6/4.0 · Highest Honors

• Concentration in Machine Learning and Enterprise Computing.
• Zell Miller Scholarship, 2018–2022.
• Math: Real Algebraic Geometry & Optimization, Algebraic Topology, Statistical Theory, Probability Theory, Real Analysis, Complex Analysis, Abstract Algebra.
• CS: Database System Implementations, Deep Learning, Machine Learning, Honors Algorithms, Intro to AI, Computer Systems & Networks.

Writing

• The t-statistic is t-distributed (2026). Proof that the OLS coefficient t-statistic follows Student-t with n−p degrees of freedom under Gaussian errors. Not peer-reviewed.
• Course and DRP notes (see Projects): The Conway Knot is not slice, algebraic cycles, abelian varieties, characteristic classes, tropical polynomials, playlist danceability, Fashion-MNIST in Julia/Flux.

Skills

• Languages and data: Python, SQL, Snowflake, Polars, DuckDB, R.
• Statistics and models: mixed-effects models, prediction intervals, experiment comparison.
• Platform: Databricks, MLflow, Tableau, Docker.
• Domain: airline revenue management, pricing, demand forecasting.