Data Scientist, AI Development

Ultimus Fund Solutions; United States; Remote; USD 120,000 - 150,000; Full time

Job summary

Ultimus Fund Solutions is seeking a Data Scientist to join an AI Development pod. You will design AI approaches for complex fund-administration workflows, shaping models, prompts, and agent logic while defining how success is measured.

You will partner with operations experts to decompose workflows, run offline and online experiments, and capture time and cost savings to guide live deployments.

Responsibilities

• Work with operations experts to break a workflow into steps and decide where AI models, AI agents, rules or people should handle each one.
• Design AI approaches for each task, including predictive models, LLM prompts, retrieval, or agent reasoning.
• Build evaluation sets and metrics (accuracy, exception rates, review time) and run offline and online evaluations.
• Analyze pilot results, find failure patterns and improve the solutions until they go live.
• Measure time and cost savings by each workflow and report results to the program team.
• Package successful patterns into reusable templates for future workflows.

Job description

About the team

Ultimus is investing in AI to transform how fund administration work gets done, from generative AI assistants, intelligent document processing and machine learning to a growing set of agentic AI workflows. Our AI Development teams are small, cross-functional groups of AI engineers, data scientists, data engineers and AI analysts who partner closely with our operations experts to redesign key workflows and build the AI solutions that support them, with AI agents an increasingly important part of that work. They are backed by a dedicated AI Infrastructure team and a shared enterprise AI platform.

About the role

As Data Scientist in an AI Development pod you design the AI approach for each workflow and prove that it works. You shape the models, prompts and agent logic, and you own how we measure accuracy, quality and savings.

What you\'ll do

• Work with operations experts to break a workflow into steps and decide where AI models, AI agents, rules or people should handle each one.


• Design the right AI approach for each task, whether predictive models, LLM prompts and retrieval, or agent reasoning, and select the right models.


• Build evaluation sets and metrics (accuracy, exception rates, review time) and run offline and online evaluations before and during pilots.


• Analyze pilot results, find failure patterns and improve the solutions until they meet the targets to go live.


• Measure the time and cost saved by each workflow and report results to the program team.


• Package what works into reusable patterns for the next workflow.