5+ years of data modeling experience across analytical and operational data systems, with significant experience designing models for large-scale data platforms (Databricks, Snowflake, BigQuery, or equivalent)
Deep expertise in dimensional modeling, Data Vault, or medallion architectures — you have strong opinions on when to use star schemas vs. wide tables vs. normalized structures, and you can justify those opinions with performance and usability tradeoffs
Production experience designing models for multi-source data harmonization — you've solved the problem of taking data from multiple vendors/systems with different schemas and creating a unified, consistent representation that downstream consumers can trust
Advanced SQL and data profiling skills — you can explore a new source system, understand its structure, identify quality issues, and design a target model in the same sitting
Experience producing Source-to-Target Mappings — you've documented complete transformation paths from source to consumption layer and worked with engineers to implement them
Strong collaboration skills with both engineering and analytics — your models need to be buildable by engineers and usable by analysts; you can speak both languages
Strong Preference
Experience with IoT, wearable, or sensor data modeling — time-series schemas for device streams, handling irregular sampling rates, modeling device-specific metadata alongside measurement data
Experience with health, fitness, or sport science data — athlete profiles, training load constructs, physiological measurements, or coaching/training plan structures
Familiarity with Databricks/Spark physical design considerations — Delta Lake partitioning strategies, Z-ordering, liquid clustering, table optimization for both batch and interactive query patterns
Experience with schema evolution and versioning strategies — managing model changes in production without breaking downstream consumers
Comfort with async collaboration across time zones — clear documentation, self-explanatory model diagrams, and proactive communication
Nice to Have
Familiarity with Nike's internal data platforms (Sole, Databricks on Nike infra)
Personal background in endurance sports, coaching, or sport science (you understand the domain because you've lived it)
Experience with data privacy modeling for health/fitness data (PII handling, consent-based access patterns)
Experience modeling for ML/AI consumption — feature stores, training datasets, or model input/output schemas
WHAT YOU’LL WORK ON
Data Architecture & Modeling
Design the unified athlete profile data model — the canonical schema that harmonizes wearable device streams (heart rate, pace, cadence, elevation, sleep, HRV, VO2max, body battery/recovery) from multiple vendors into a single, query-optimized representation on Databricks/Sole
Model the coaching intelligence data layer — defining how structured inputs (intake questionnaires, training goals, fitness baselines, injury/pain history) and unstructured inputs (coach notes, athlete feedback, session RPE) are organized, linked, and exposed for consumption by the AI Coaching Engine and coaching dashboards
Design the NSRL-to-athlete data model — the schema that translates Nike Sport Research Lab's sport science constructs (Horwill's predictors, Daniels VDOT, Acute Chronic Workload Ratio, training periodization models) into production-ready data structures that coaching algorithms can consume at scale
Own the Bronze/Silver/Gold medallion layer design for the Activity domain — defining the transformation contracts, grain, and business rules at each layer within the MAC BI standardized pipeline template
Source-to-Target Mapping & Integration
Produce Source-to-Target Mappings (STMs) for all Activity data sources — documenting the complete transformation path from raw device vendor schemas through to Gold-layer consumption models
Handle the modeling complexity of multi-vendor device data — resolving schema conflicts, mapping equivalent metrics across devices (e.g., Garmin Body Battery vs. Apple Watch recovery estimates vs. COROS stamina), defining rules for when sources overlap or conflict
Design data contracts between the Activity data foundation and downstream consumers — defining the API schemas, refresh cadences, SLAs, and versioning strategy that NRC/NTC product teams, the AI Coaching Engine, and coaching staff tools depend on
Model race and sport experience data — training plan structures, race-specific schemas (5K through ultra), in-race coaching signal formats, and post-race performance analysis structures across NRC marathon programs
Data Quality & Governance
Define data quality rules at the model level — embedding Spark Expectations checks, referential integrity constraints, and business rule validations directly into the model design so quality is structural, not an afterthought
Establish the Activity data dictionary and catalog — authoritative definitions for every entity, attribute, metric, and relationship in the Activity domain, maintained as a living artifact that evolves with the platform
Own the schema evolution strategy — how do the models handle new device vendors, new sensor types, new sport science models, and new coaching features without breaking existing consumers?
Partner with data analysts to validate that model designs support the analytical use cases — ensuring the metric layer, dashboards, and ad-hoc analyses can be served efficiently from the physical model
What Success Looks Like (First 12 Months)
The unified athlete profile model is in production — wearable device data from multiple vendors flows through a single, well-documented schema that downstream systems depend on
Source-to-Target Mappings are complete for all Activity data sources and actively used by engineers building pipelines
The coaching intelligence data model supports both the AI Coaching Engine and coaching dashboards without requiring separate transformations for each consumer
The Activity data dictionary is established and adopted as the single source of truth for what the data means
Schema evolution processes you designed have handled at least one new device vendor or sensor type without breaking existing consumers
Data quality rules embedded in the model catch issues at ingestion time, not after they reach dashboards
We offer a number of accommodations to complete our interview process including screen readers, sign language interpreters, accessible and single location for in-person interviews, closed captioning, and other reasonable modifications as needed. If you discover, as you navigate our application process, that you need assistance or an accommodation due to a disability, please complete the Candidate Accommodation Request Form.
