Senior Machine Learning Engineer

Steadily•Austin, TX

6d•Hybrid

About The Position

As the second engineer joining our dedicated ML team, you will add input on our architecture, tooling, and ML strategy. Because we are a fast-growing, agile company, this is a true end-to-end role. You’ll own building, evaluating, and deploying your models to production, as well as monitoring them for quality and accuracy over time. We operate across data types including public, proprietary, and a large volume of image data. Because we currently operate without a dedicated Data Engineering team, you will also own the data layer for your models. In practice, that means you'll often be the first person to touch a given raw data source; you're comfortable going from an unrefined, previously unexplored data set through feature engineering and into a production ML model. You can expect roughly a 30% data pipeline / new dbt table building (lightweight, not heavy ETL) and 70% feature engineering, modeling, deployment, and monitoring split in your day-to-day work. You’ll operate with a high degree of autonomy and serve as a trusted technical owner for business problems across the organization. Steadily is still early in our exploration of where AI/ML models can drive the biggest value, making this role ideal for engineers who thrive in ambiguous environments and want their technical work to translate directly into massive business impact.

Requirements

5+ years experience applying Machine Learning methods to production problems.
Able to dive into a complex codebase without too much spin-up.
Full-stack with data: Comfortable starting from a raw, unrefined data source that no one has previously worked with, building the lightweight pipeline or dbt table to make it usable, and carrying it all the way through feature engineering, modeling, and deployment.
Builder with a Business Mindset: Likes the product-side of data and thinks about how to apply modeling and evaluation techniques to real-world problems. Has thoughtful opinions about where the data leads and how to maximize the business impact of your work.
Pragmatic: Prioritizes impact and delivery. Balances speed and quality, making thoughtful trade-offs to solve problems effectively. Leverages off-the-shelf solutions (and foundational models) so we don’t reinvent the wheel, but understands when a custom solution is appropriate.
Curious: Not just an order-taker. Curious about what makes the business tick and learns the intricacies of how it runs. This results in strong intuition for when an analysis is wrong and leads you to suggest ideas and insights that nobody thought to ask for. Not the type of engineer who wants fully fleshed-out specs thrown over the wall for you to implement.

Nice To Haves

Actuarial experience, or experience applying models to risk evaluation and aggregation problems.
Experience in computer vision and image analysis.
Experience with dbt or similar modern data transformation tools.
Past experience as a team lead or owning end-to-end deployment is definitely a plus.

Responsibilities

Own the end-to-end ML lifecycle: Design, build, deploy, and evolve data sets and models with an emphasis on scalability, quality, and maintainability. Focus areas could include estimating property-level risk, accurately assessing costs, and using aerial image analysis or modeling techniques to identify attributes that feed into other models.
Build and maintain the data layer: Build lightweight data pipelines and new dbt tables to get raw data model-ready, without owning heavy ETL infrastructure.
Drive measurable business impact: Lead the exploration and implementation of new ML applications in our product ecosystem to better predict risk on a per-insured level and in aggregate across the entire portfolio.
Write clean, maintainable code in our stack with the help of AI: We build on an event-driven architecture using Kafka, AWS (EKS), Python, Django/FastAPI, and Postgres, with a full CI/CD pipeline via GitHub Actions. You will set a high bar for engineering quality and architectural design within this ecosystem.
Partner closely with Engineering, Product, Operations, and Business teams to design reliable solutions across systems and ensure your models are solving real-world problems.
Provide excellent metrics and visibility into model quality, bias, and performance to assess how it’s helping the business, ensuring a high bar of scientific rigor and evaluation.