Data Scientist (Databricks, Mosaic AI)

About Peregrine Peregrine Advisors is a rapidly growing AI and data innovation hub helping government solve hard problems and deliver better outcomes. We work across the full lifecycle—from strategy and governance through architecture, engineering, analytics, AI, visualization, and training—for clients including the SEC, PCAOB, DDOT, and DCAA.

Our clients describe our support as “absolutely excellent” and our people as “instrumental to organization’s success.” Winner of the United Nations COP27 Datathon, our work has earned all Exceptional CPARS ratings, SEC Chair’s Awards, the SEC Vision Award, and recognition from FDA leadership.

We are also built differently. Peregrine is the first and only social enterprise organized as a Benefit Corporation and SBA 8(a) company focused on AI and data for government. We believe business success should create client value, social value, and opportunity for our people.

As we enter our next stage of growth, we need dependable, motivated, and skilled people who have a combination of intellectual curiosity and intellectual courage. With new clients, larger programs, and expanding responsibilities, the people joining Peregrine now will have the opportunity to help shape the company we become.

People first. Ownership always. We solve problems through people, process, and technology, in that order. Technology matters, but people make transformation happen. That principle defines our culture. Trust and motivation are the foundational pillars of delivery on which our skills stand. We hire people we can trust, who care deeply about what they do, take ownership, learn quickly, help one another, and refuse to walk past a problem because it belongs to someone else.

At Peregrine, you are not joining simply to fill a position. You are joining a team, taking ownership of a mission, and helping build a company. If that excites you, we should talk.

At Peregrine Advisors, you will build, evaluate, and put into production the machine learning and generative AI models a federal agency relies on. But we hire people, not seats. As the work evolves, you will learn new systems and tools, take on greater responsibility, and help develop the firm’s capabilities, tools, and lines of business. We move our best to where the hardest problems are.

This is a full-time W-2 position with a salary of $110,000 to $140,000 per year. It is a remote position. The initial engagement requires United States citizenship; dual citizenship and green card holders are prohibited by the contract terms. It also requires the ability to satisfy the required suitability, security, and access requirements.

Your first project Your first project will likely have you taking models and AI agents from experiment to production on Azure Databricks, with the evaluation evidence to show they work. Depending on the assignment, you may:

• Develop and train machine learning models on Azure Databricks, tracking experiments, models, and versions in MLflow
• Build reusable features in the Databricks Feature Store, and serve models through Databricks Model Serving
• Build retrieval and agent applications with Agent Bricks and Databricks AI Search (formerly Vector Search), using Azure OpenAI models through Microsoft Foundry
• Evaluate model and agent quality with MLflow evaluation, including large language model (LLM) judges, labeled test sets, and human review
• Monitor models in production for drift and quality, and write the documentation a model review asks for

You deliver models the agency can trust and explain. You measure quality before and after release, record what each model is for and where it fails, and leave the team pipelines it can retrain and rerun without you.

This is where you start, not the shape of your career here.

Requirements

• 4+ years of data science or machine learning engineering experience, including models deployed to production
• Bachelor’s degree
• Python for data science and machine learning, including pandas, scikit-learn, and PySpark
• Hands-on Azure Databricks work, including MLflow for experiment tracking and model management
• Deploying and monitoring models in production, including model serving, versioning, and drift monitoring
• Building LLM applications, including retrieval-augmented generation (RAG)
• Evaluating model and LLM output against defined metrics, test sets, and human review
• Structured Query Language (SQL) and …