Data Scientist ll

Location: Kirkland, WA HQ

Department: Technology

Location Type: HYBRID

Employment Type: FULL_TIME

• Assess and structure training data: Evaluate large-scale datasets for quality, apply structured labeling frameworks, and diagnose systemic gaps or inconsistencies that affect model performance
• Train, update, and apply models against existing data: Train or fine-tune models, apply scoring/classification tools against datasets to auto-tier or flag quality at scale, and iterate on models as new data or requirements emerge
• Surface insights: Explore labeling and usage patterns to find where data quality is strong or weak, and turn findings into clear, defensible recommendations
• Support data pipeline integration: Partner with engineers to ensure clean, well-labeled data flows smoothly into downstream systems and model training workflows
• Test and validate: Validate model outputs and quality-scoring results against ground truth before they're used downstream, and monitor for drift or regressions as data and models evolve
• Contribute to tooling improvements: Provide input on internal tools used for data labeling and review, helping shape what they need to track, automate, or support
• Build quality frameworks: Define metrics and benchmarks that answer “how good is our data?” and track them over time, with a direct eye toward how data quality impacts model performance

• Master's degree in Data Science, Statistics, Computer Science, or a related quantitative field
• 1–4 years of experience in data science, machine learning, or a related quantitative field
• Solid understanding of machine learning fundamentals — how training data quality affects model performance, common failure modes from bad/biased data, and how to evaluate a model's output critically
• Hands-on experience training, fine-tuning, or updating ML models — not just applying pre-built tools or doing descriptive analysis
• Strong SQL and Python skills (pandas, numpy, scikit-learn, or a deep learning framework such as PyTorch/TensorFlow); comfortable exploring and cleaning large, messy datasets independently
• Practical experience working with AI coding tools (e.g., Copilot, Cursor, Claude Code, or similar) as part of your daily workflow, with an agentic mindset — comfortable delegating well-scoped subtasks to AI tools/agents, verifying their output, and iterating quickly rather than hand-coding everything from scratch
• Ability to translate ambiguous questions into measurable, trackable metrics
• Familiarity with data labeling or annotation workflows is a plus, not required
• A quality-first, curious mindset — interested in why data issues occur, not just flagging them
• Exposure to computer vision or video/image datasets specifically is a plus

• Measurable improvement in data quality and coverage over time
• Clear, trusted quality benchmarks that inform data and modeling decisions
• Strong collaboration with engineering on data pipeline health