Aaron ("AJ") Steers

Experienced architect, data and AI engineer, software developer, and technical leader. I am passionate about AI, data, and open source software. I have designed robust end-to-end data systems at Amazon scale, and I have created several open source projects related to data engineering and data transformation. I regularly contribute to other open source projects, especially those that I can use personally or that my customers are using. I am an experienced manager but I'm also a deeply technical IC. I thrive in both IC and Manager/Director roles, preferring to focus where ever the need is greatest for an organization. I also have experience manager AI agents such as Devin.ai - having merged over 100 AI-authored pull requests. My most successful open source projects: - Meltano SDK - https://github.com/meltano/sdk - A framework for building data connectors, currently has over 80 contributors and is used by over 1,300 repositories on GitHub. - PyAirbyte - https://github.com/airbtehq/PyAirbyte - Allows Airbyte data pipelines to be run anywhere, including Jupyter and Google Colab notebooks. Currently has 5K downloads per week. - DataOps Infrastructure Catalog (https://github.com/slalom/dataops-infra, Archived) - A Terraform repository of over a hundred re-usable infrastructure components, aimed to help data teams quickly and easily deploy cloud infrastructure for their data plane. Specialties: - Building quality SDKs that developers can use to deliver best-practice data solutions with minimal effort. - Directing and managing both human programmers as well as AI workforce resources such as Devin.ai. - Designing cloud data systems that scale up to massive datasets, while also being able to scale down to near-zero cost when appropriate. - Implementing AI and LLM solutions, including: parsing unstructured data, feeding vector stores, building agents and RAG pipelines, etc. - Giving technical talks at conferences and meetups, providing training to technical and non-technical audiences alike. Approved US Patents: - View-based data mart management system (US #10846284) - A predecessor to dbt (data build tool), submitted approximately 18 months before dbt launched. Defines data pipelines as SQL "SELECT" statements, which can be deployed anywhere. - Distributed data validation service (US #10248508) - A novel declarative testing framework which can run well in distributed environments.