Applied Data Scientist - Analytics & AI Agents

About Alpha Nodus

Alpha Nodus builds Gravity, an AI-powered operations platform for outpatient diagnostic imaging centers and radiology practices in the U.S. Gravity's agents handle work that imaging centers still do by hand: reading faxed orders, checking insurance, getting prior authorizations, scheduling patients and chasing payment. Every one of those workflows produces data about how the practice runs and how well our agents perform.

The role

You'll be the person who understands that data better than anyone. You'll make our AI agents measurably better, turn operational data into insights our customers act on, and tell engineering exactly what to capture so both are possible. You'll work across product, engineering and customer success, and you'll own the question "how do we know it's working?"

What you'll do

Make our agents better (about 50%)

• Build and maintain evaluation sets for each agent: real cases with known correct outcomes, refreshed as workflows change.
• Read agent runs closely, group failures into a clear taxonomy, and size each failure type by frequency and impact.
• Diagnose root causes: prompt, context and retrieval, tool behavior, upstream data quality or a gap in the workflow itself.
• Propose and test fixes with engineering, compare agent versions on your eval sets and in production, and report the result in numbers.
• Track agent health over time: accuracy, automation rate, human-override rate, turnaround time and cost per task.

Find insights customers care about (about 30%)

• Analyze operational data across customers to answer questions imaging center leaders ask: where authorizations stall, which payers deny most and why, where schedule capacity sits unused, where revenue leaks.
• Turn analyses into short, decision-ready findings for customer reviews and product decisions.
• Define the core metrics Gravity reports to customers, with clear definitions everyone uses the same way.

Shape the data foundation (about 20%)

• Design analytics-ready data models: the right grain, fact and dimension tables, event logs and history tracking.
• Write precise specs for engineering: which fields, in which tables, at what grain, populated by which events.
• Define what agents should log (inputs, tool calls, outputs, confidence, human corrections) so every run can be analyzed later.
• Handle patient data responsibly: de-identify where possible, follow least-privilege access and keep PHI out of places it doesn't belong.

What you'll bring

• 5–8 years in data science, product analytics or analytics engineering, including time at a startup or on a small team.
• Expert SQL and strong data modeling: you think in grain, joins and history, and you can explain why a table should be shaped a certain way.
• Fluent Python for analysis, plus enough statistics to tell a real change from noise.
• Hands-on work evaluating or improving LLM-based applications: building eval sets, doing error analysis, and testing prompt or context changes.
• A habit of reading raw examples before trusting a dashboard.
• Clear writing. You can produce a one-page finding for a customer and a precise schema spec for an engineer.
• Comfort pushing back with evidence and influencing people you don't manage.

Nice to have

• U.S. healthcare data experience: HL7 (ORM, ORU, SIU), FHIR, X12 EDI (270/271, 278, 835, 837), payer behavior or revenue cycle.
• Experience in radiology, imaging, or another operations-heavy healthcare setting.
• Work with HIPAA-regulated data.
• Experience with dbt, event tracking design or LLM tracing and observability tools.
• Presenting findings directly to customers.

Why this role

• Your work shows up directly in product quality and in what customers see.
• You'll set how a growing company measures its AI and its value to customers.
• Small team, real ownership, and a problem space (U.S. imaging operations) where good data changes how patients get care.