Lead Data Scientist
(Risk Modeling / Python / AWS / SageMaker)
Polaris I/O — Remote (U.S. only) | Full-time
About the company
Polaris I/O provides the world’s only integrated go-to-customer platform that combines transformation services, industry insights, and powerful technology to enable B2B companies to protect, retain, and grow large accounts. Our platforms take an outside-in approach to improving commercial health, utilizing executive buyer insights to inform the orchestration of commercial activities that improve relationships, maximize growth, and scale beyond our engagement.
Role Summary
Polaris I/O is hiring a Staff Data Scientist to lead the design, validation, and ongoing health of the scoring models that power our risk and analytics platform. Our system ingests structured client data, news content, and public data sources to produce risk, threat, and anomaly scores across a range of business domains — with vendor and supply chain risk among the first use cases.
This is a foundational hire. You will be the first dedicated data scientist on the team and will own the model side of the platform end-to-end — from feature design and dataset preparation, through model authoring and validation, to deployment, monitoring, and iteration. You will partner closely with the Data Engineering team that builds and operates the underlying data platform, and with the CTO and product leadership on model strategy.
Our scoring architecture is intentionally model-agnostic, so you will have meaningful influence over which models we build, how they are configured, and how we measure success. The platform is designed to support both formula-based models and service-based models deployed as endpoints.
Our stack includes:
- Python (pandas, scikit-learn, statsmodels, NumPy)
- AWS — SageMaker, S3, Aurora PostgreSQL
- Jupyter / SageMaker Studio for exploratory and validation work
- YAML-driven model and feature configuration
This role is hands-on. You will write code, run validations, and own model quality from prototype through production.
What You’ll Do
Model Development
- Design, train, and validate scoring models across a range of risk domains
- Define features, weights, thresholds, lookback windows, and data handling rules in our model configuration format
- Build both formula-based models and service-based models, choosing the right approach for each problem
- Generate reproducible training, validation, and test datasets and document the splits clearly enough for any teammate to reproduce them
- Partner with engineering on the handoff from model development to production deployment
Model Validation & Governance
- Establish validation standards and a governance framework for the platform — thresholds, cohort tests, and the criteria a model must meet before it ships
- Run validation studies with the right metrics for the model type (e.g., AUC-ROC, precision/recall, calibration error, true/false positive rates, detection latency) and break results down by relevant cohorts
- Author and own recurring model health reports — covering drift, cohort performance, and score distribution
- Diagnose model failures in production and drive remediation
- Maintain reproducibility — every validation result should trace back to a specific dataset and a versioned model configuration
Model Lifecycle & Operations
- Manage models through their full lifecycle, including planned replacements and re-scoring of affected entities
- Define retraining cadence and triggers based on model health
- Produce explainability outputs (score rationale, summary explanations) suitable for client-facing surfaces and audit
- Translate new modeling needs into requirements that the data engineering team can build into the underlying data pipeline
Cross-Functional Leadership
- Serve as the data science point of contact for product, engineering, and platform teams
- Influence the long-term direction of our model-agnostic scoring architecture
- Mentor engineers and analysts on modeling fundamentals, validation rigor, and reproducibility
Required Job Qualifications
- 10+ years of professional experience in data science, statistical modeling, or applied machine learning
- Demonstrated experience designing and shipping production models in at least one risk-relevant domain: vendor risk, supply chain risk, threat assessment, fraud, anomaly detection, or comparable scoring/classification problems
- Expert-level Python for data science work — pandas, scikit-learn, statsmodels, NumPy
- Strong SQL skills, including the ability to write performant queries against PostgreSQL/Aurora for feature extraction and validation
- Deep expertise in at least one full statistical environment for exploratory analysis and validation: Python (Jupyter), R, SAS, or SPSS
- Production experience with model validation methodology — train/validation/test splits, cohort-based validation, calibration, drift detection, and metric selection
- Experience deploying models as services (REST endpoints, SageMaker, or equivalent) and reasoning about latency, scaling, and operational behavior
- Working knowledge of AWS — S3 is required; familiarity with SageMaker, Aurora, and IAM is expected
- Strong written and verbal communication in English — this role includes model documentation, validation reports, governance frameworks, and collaboration with U.S.-based teams and clients
Preferred Qualifications
- AWS SageMaker production experience (SageMaker Studio, model endpoints, model registry)
- Time series and anomaly detection methods — trend analysis, seasonality, spike detection, peer benchmarking
- Experience working with text- or sentiment-derived features (fine-grained sentiment, emotion-based sentiment, NLP-extracted entities)
- Experience defining and operating formal model governance frameworks
- Familiarity with deep learning frameworks (PyTorch, TensorFlow)
- Experience working alongside Python/AWS engineering teams in an event-driven microservice environment
Work Authorization Requirement: Candidates must be legally authorized to work in the United States on a full-time, ongoing basis without the need for current or future employer sponsorship (for example, we are not able to sponsor employment visas now or in the future)Work Expectations
Work Hours: This is a full-time role that requires you to be online and working during standard business hours, with occasional after-hours support based on business needs.
Remote Work: This position is fully remote and all work must be performed within the United States. Candidates must be legally authorized to work in the United States. Regular working hours will align with the time zone.
Conflict of Interest: You are expected to devote your full professional time and attention to Polaris I/O and not take on other employment or contract work that interferes with your responsibilities, competes with the company, or creates a potential conflict of interest.
Compensation & Benefits:
Polaris I/O offers a competitive salary, a comprehensive benefit package including medical, dental, and vision coverage, a generous PTO plan, and opportunities for professional growth and development.
Polaris I/O is an Equal Opportunity Employer. We consider all qualified applicants without regard to any characteristic protected by applicable federal, state, or local law.
Pay: $170,000.00 - $200,000.00 per year
Benefits:
- Dental insurance
- Health insurance
- Vision insurance
Application Question(s):
- Do you have expert-level Python for data science work — pandas, scikit-learn, statsmodels, NumPy?
- Do you have demonstrated experience designing and shipping production models in at least one risk-relevant domain?
- Do you have deep expertise in at least one full statistical environment for exploratory analysis and validation: Python (Jupyter), R, SAS, or SPSS?
- Do you have experience deploying models as services (REST endpoints, SageMaker, or equivalent) and reasoning about latency, scaling, and operational behavior?
- Are you comfortable owning the full model lifecycle end to end?
Experience:
- in data science, modeling, or applied ML: 10 years (Required)
Work Location: Remote
