Steve Wilson

About Steve Wilson

I work at the intersection of specialized and generative AI. I build models that forecast…

Experience

1. Principal Ai Architect

   Cirium

   Aug 2016 — Present · Chapel Hill, NC, US

   Shipped DeepETA, a deep learning system for real-time flight delay and ETA forecasting. Achieved state-of-the-art accuracy across global traffic flows, outperforming legacy and vendor models. Deployed via Databricks Model Serving for high-availability inference at scale.Designed graph neural networks to analyze how delay propagates through airline and airport networks. Created interpretable, optimization-ready models using PyTorch, DGL, cvxpy, and Spark, enabling targeted buffer placement and delay mitigation.Led development of Cirium EmeraldSky an AI-powered carbon emissions platform used by top airlines. Built automated Spark pipelines with historical aircraft and seating data to deliver the most precise per-flight emissions estimates in the industry.Built discrete event simulations of airport gate and runway operations using SimPy and Pandas. Used to evaluate schedule robustness, identify throughput constraints, and recommend operational improvements under peak demand scenarios.Prototyped aviation intelligence assistants using LLMs, LangChain, Text2SQL, and Prompting Strategies (Few-Shot Learning, CoT, Reflection). Delivered natural language interfaces for querying structured data, allowing analysts and planners to extract insights without writing code.Architected foundational ML pipelines across flight movement, weather, and tracking data. Used Databricks Asset Bundles, AWS, and Spark to create reliable, scalable, and reusable pipelines that support both R&D and production-grade inference.

Education

• Boise State University

  BBA, Finance

  1996

• University of Phoenix

  Master of Business Administration (MBA), Accounting

  2004 — 2006