Data Scientist - Decision Intelligence & Healthcare ML

About the Role

We're a healthcare technology company building tools that help provider organizations, ACOs, and health systems make better financial and operational decisions. Our platform combines data engineering, machine learning, and agentic AI to turn complex healthcare data into actionable intelligence.

As a Data Scientist, you'll own the ML models and statistical systems that drive the platform's intelligence layer. You'll build models across a range of domains — financial forecasting, payment integrity, population health surveillance, provider benchmarking, and more. Your work feeds directly into agentic AI workflows that translate model outputs into plain-language narratives and recommended actions for end users.

This is applied ML in a domain where precision matters — every false positive costs an adjudicator's time, every missed forecast erodes trust, and every accurate recommendation recovers real dollars or prevents real losses.

What You'll Do

Decision Intelligence Models

* Build forecasting models for financial planning: revenue projections by payer/contract/service line, cost trend forecasting, TCOP (total cost of patient) projections, budget variance prediction, and capitation rate modeling

* Design and maintain benchmarking algorithms: peer cohort construction from multi-dimensional provider/org attributes, percentile distribution computation, CMS national benchmark integration, and outlier detection for performance management

* Build visit optimization models: slot utilization analysis, no-show prediction, revenue-per-visit optimization, scheduling pattern analysis, and capacity planning

* Develop VBC performance scoring: quality measure forecasting, shared savings projections, risk adjustment optimization (CMS-HCC), attribution modeling, and contract performance simulation

Payment Integrity Pipeline

* Build and maintain detection models covering DOFR routing misdirection, duplicate detection (LSH), CC/MCC/SOI severity inflation (Isolation Forest + XGBoost), procedure code integrity (NCCI rules + XGBoost), discharge/transfer violations (PACT rules + SQL lookback), and contract rate misapplication (range checks + CUSUM drift)

* Calibrate dollar-weighted work queues (P1/P2/P3 tiers) to surface high-value flags, optimizing for precision over recall

* Design feedback loops: confirm/clear workflows with structured reason codes, dollar impact tracking, monthly model retraining, weekly signature updates, and quarterly threshold recalibration

Population Surveillance & Anomaly Detection

* Implement multi-tier population surveillance systems: descriptive statistics, CUSUM control charts, peer group z-scores feeding heightened scrutiny lists, and time-series forecasting with anomaly detection

* Monitor model and population drift via statistical process control, triggering retraining when sustained shifts are detected

* Build anomaly detection across multiple data domains — not just claims, but also revenue cycle metrics, scheduling patterns, GL variances, and operational KPIs

Explainability & Agentic AI Integration

* Build SHAP-based explainability for all ML models and work with the AI agent team to synthesize plain-language decision narratives — "what happened, why, what it costs, what to do"

* Maintain vector-embedded error signature libraries for semantic matching of known patterns against new data

* Design model outputs specifically for agentic consumption — autonomous AI workflows call your models via tool use, so output schemas, confidence calibration, and actionability scoring directly impact agent quality

* Contribute to decision intelligence prompts: help design the reasoning chains that translate model outputs into recommended actions

Working with AI

* Use AI-assisted development tools daily for model prototyping, feature engineering, code generation, and analysis

* Understand agentic AI patterns — your models are consumed by autonomous agents, not just displayed on dashboards. You'll design for tool-use invocation, multi-step reasoning chains, and human-in-the-loop review workflows

* Collaborate with the engineering team on agent action groups, ensuring model outputs are structured for agentic consumption

What We're Looking For

Required:

* 3+ years of applied ML experience, with production model deployment (not just notebooks)

* Strong experience with gradient boosting (XGBoost/LightGBM) — training, hyperparameter tuning, feature engineering, SHAP interpretation

* Experience with anomaly detection: Isolation Forest, CUSUM/statistical process control, or similar unsupervised methods

* Experience with time-series forecasting (Prophet, ARIMA, or similar) for financial or operational projections

* Proficiency in Python (scikit-learn, XGBoost, pandas, numpy) and SQL for feature engineering against large datasets

* Experience with model evaluation in precision-critical settings — understanding precision/recall tradeoffs, threshold calibration, capacity-constrained optimization

* Familiarity with vector embeddings and similarity search (sentence-transformers, FAISS, pgvector, or similar)

* Comfort with AI-assisted development — you use AI tools daily for prototyping, analysis, and code generation

* Understanding of agentic AI patterns — you know how autonomous agents consume model outputs and can design for that consumption pattern

* Ability to communicate model behavior to non-technical stakeholders — you'll work closely with finance and operations leaders

Preferred:

* Healthcare domain experience: claims data, DRG groupers (MS-DRG, APR-DRG), NCCI edits, CMS payment policies, ICD-10, CPT/HCPCS coding, VBC contracts, risk adjustment

* Experience with multiple healthcare data types beyond claims: EMR/clinical data, revenue cycle metrics, financial/GL data, scheduling/operational data

* Experience building ML systems consumed by AI agents or autonomous workflows (tool-use patterns, structured output design, confidence calibration for agentic consumption)

* Experience with LSH / MinHash for near-duplicate detection

* Familiarity with model registries (MLflow or similar)

* Experience with cloud-based ML pipeline orchestration (SageMaker, Step Functions, Vertex AI, or similar)

* Exposure to HIPAA/BAA compliance requirements for ML systems handling PHI

* Knowledge of CMS PACT transfer policy, NCCI bundling rules, or physician fee schedule RVU calculations

Tech Stack

* ML Frameworks: XGBoost, LightGBM, scikit-learn, Prophet, datasketch (MinHash LSH)

* Explainability: SHAP, sentence-transformers

* Vector Store: pgvector or similar

* Data: Cloud data warehouse (Redshift, BigQuery, Snowflake, or similar), PostgreSQL

* Orchestration: Cloud workflow orchestration (Step Functions, Airflow, or similar)

* Model Registry: MLflow or equivalent

* AI Integration: LLM platform (Bedrock, Azure OpenAI, or similar) — your models feed agentic decision workflows

* Languages: Python, SQL, TypeScript

* Development: AI-assisted coding tools

Compensation

Competitive salary commensurate with experience, equity, and benefits.