AI Cloud Data Scientist

UMB’s Data Science team partners directly with business and technology leaders to turn high-value banking problems into governed, measurable AI solutions—agentic AI systems, LLM applications, and machine learning models that automate and modernize how the bank works.

As an AI Cloud Data Scientist, you will decompose business processes with line-of-business partners, determine whether the right answer is an AI agent, a machine learning model, a deterministic rule, or a combination, and build the solution far enough to prove its value. You will also build the reusable tools, patterns, and evaluation practices that make the broader data science team more capable, working closely with a Principal Data Scientist and partner engineering teams.  This role owns problem framing, solution design, prototyping, evaluation, and production-ready handoff. Partner engineering teams own production deployment, scaling, and operations unless otherwise assigned.  Overall responsibilities which will include multiple initiatives assigned by Data Science and IT leadership.

This role is hybrid (Mon through Thu on-site / Fri remote) for candidates in the Kansas City metropolitan area and open to qualified remote candidates outside of the Kansas City area.  However, the remote location must be within the US.

How you’ll spend your time:

• Agentic AI & LLM Solution Development:
  • Design and build multi-step agentic AI systems—tool and function calling, orchestration, retrieval, and memory—on AWS Bedrock Agents or an equivalent platform.
  • Engineer prompts and context, including the business and semantic context agents need to accurately answer natural-language questions from the business.
  • Prototype, iterate, and prove solutions in a development environment, and apply large language models to automate, augment, and replace manual workflows.
• Business Translation & Partnership:
  • Partner with business teams to decompose processes, identify outdated systems and manual work that are strong candidates for AI modernization, and choose the simplest method that solves the problem well.
  • Solve the underlying problem rather than putting a patch on it, own the solution from discovery through validated results, and communicate tradeoffs and outcomes clearly to business and technical audiences.
• Machine Learning, Evaluation & Governance:
  • Build, evaluate, and reason about machine learning models end to end when the problem calls for it, applying sound experimentation, feature engineering, and validation practices.
  • Build evaluation harnesses and define metrics to measure agent and model quality, including failure analysis and regression testing.
  • Implement guardrails, human-in-the-loop checks, and the security, compliance, and governance controls a regulated banking environment requires.
• AI Enablement & Handoff to Engineering:
  • Build reusable skills, workflows, tools, and patterns—for prompts, context, tool interfaces, evaluations, and approvals—that improve data science productivity and quality across the team, and share them through examples and mentoring.
  • Define data, pipeline, and integration requirements, and hand off solutions to engineering with documented designs, evaluation results, and acceptance criteria.

In your first year, success looks like a high-friction bank process turned into a governed agentic workflow with measurable value, and a reusable evaluation approach that other initiatives adopt.

We’re excited to talk if you have:

• At least 3 years of experience across data science, machine learning, AI engineering, applied research, or enterprise software development, plus a Bachelor’s degree in a related field or an equivalent combination of education and experience. We value demonstrated ability and project depth over an exact title or degree path.
• Strong proficiency in Python and SQL, solid machine learning skills, and work with normal engineering discipline: Git, testing, documentation, and APIs.
• Built at least one multi-step LLM workflow involving tool or function calling, orchestration, retrieval, structured outputs, state, or human approval paths.
• Ability to evaluate AI and model outputs using repeatable tests, meaningful quality metrics, and failure analysis.
• Ability to translate ambiguous business problems into AI solutions, scope them with stakeholders, and explain tradeoffs to non-technical audiences.

Bonus Points if you have:

• Built …