Staff Data Scientist

About this role

Purpose Unlimited is an independent financial services company with an unrelenting focus on customer-centric innovation, delivered through technology-driven solutions. Led by entrepreneur Som Seif, the company is developing a diversified product platform aimed at addressing historically underserved segments of the market. Purpose Unlimited’s businesses include Purpose Investments, Advisor Solutions by Purpose, and Driven.
Vacancy Status: This is for a current vacancy
Job Description Who you are Customer behavior is a signal. Most companies collect it. Few know how to read it. You will be one of the early data science hires shaping how Purpose understands the clients on our wealth platform — and how the platform responds to what we learn. You'll build the measurement architecture, behavioral models, and causal infrastructure that connects product decisions to client outcomes. Not what looks like it works. What actually does. This is a foundational role at a foundational moment. The intelligence layer you build will shape how Purpose designs, measures, and evolves its platform for years — and increasingly, how our AI systems understand and serve the clients who depend on us. You're an engineer as much as a scientist — you own the work from raw event data to deployed system, and you hold yourself accountable for whether it changed how the business operates. What AI does so you don't have to

• Recurring insight generation — automated analytics workflows run on schedule with defined guardrails. You design the system and own the governance; you don't write the weekly summary.

• Boilerplate feature engineering — AI coding assistants accelerate pipeline construction from event data. You define what to build and review what the tools produce.

• First-pass segmentation refreshes — classification and scoring jobs re-run automatically. You set the standards and monitor for drift, you don't rerun the job.

• Standard visualization and reporting — AI tools increasingly generate monitoring views from structured data. You define what signals matter; the generation is handled.

What only you can do

• Determine what is worth measuring — in a platform generating thousands of behavioral signals, the most valuable judgment is deciding which three metrics actually reflect client value and which are vanity.

• Push back on attribution that flatters — AI systems generate correlations at scale. You are the person who asks whether the relationship is causal, whether the experiment is valid, and whether the conclusion will hold when the business acts on it.

• Translate behavioral reality into a measurable operating model — starting from how clients actually make financial decisions, not from how the product team assumes they do.

• Design AI analytics systems with the rigor of a statistician — evaluating LLM-generated insights with the same calibration and drift standards you apply to your own models.

• Build the institutional trust that makes product and business leaders act on quantitative findings, including findings they didn't want.

What you will own

• The client intelligence architecture: the measurement model that maps how clients discover, onboard, activate, and engage with Purpose's wealth platform — with FOMs that connect platform activity to real client outcomes.

• Behavioral models that change decisions: client segmentation, lifecycle classification, churn prediction, propensity scoring, and behavioral clustering — validated against real outcomes and governed for drift.

• Causal measurement with discipline: experiment design and analysis, causal inference when experiments aren't possible, and a consistent standard for what counts as evidence that a product or AI intervention actually worked.

• Agentic analytics governance: you set the standards for what AI-generated insight requires human review, what guardrails govern automated outputs, and how the intelligence layer stays accurate and auditable in a regulated environment.

What you must bring

• Deep fluency in Python and SQL — owning work end-to-end from raw event data to deployed software, with production-grade testing and documentation standards.

• Proven experience building and evaluating predictive models in production-like settings: classification, survival models, segmentation, scoring, calibration, and cost-sensitive decision frameworks.

• Applied causal inference — A/B testing, matching/propensity, diff-in-diff, regression discontinuity. You know the difference between a valid experiment and a flattering one.

• Experience in a regulated industry (financial services, healthcare, or equivalent) — you treat data access controls, model auditability, and responsible use of customer data as core professional standards.

What will set you apart

• You've been early on a data science or analytics team and built measurement and modeling foundations before the playbook existed — you know what it means to make something from scratch that others depend on.

• You have experience evaluating LLMs or AI-generated outputs with statistical rigor: calibration, drift detection, bias assessment. You apply the same skepticism to AI outputs that you apply to your own models.

• You've used AI-assisted development tools (GitHub Copilot, Claude Code, Cursor, or equivalent) as a genuine productivity multiplier in data science workflows.

• You have applied behavioral economics or behavioral finance principles to how you designed metrics or interpreted client behavior — you think about why clients make financial decisions, not just what they do.

What success looks like in your first year

• A measurement architecture is live — FOMs that connect platform activity to client outcomes are defined, instrumented, and being used by product and business leaders to make decisions they couldn't make before.

• At least two behavioral models are in production — churn prediction, propensity scoring, or lifecycle classification — with monitoring, documentation,…