Senior Data Scientist, Evolv Technology

• Master’s or PhD in Data Science, Computer Science, Applied Mathematics, Statistics, Physics, or related field
• 2-3+ years of data science experience working with real‑world ML datasets (time‑series, images, video, sensors)
• Proficiency in Python and data science libraries (NumPy, pandas, matplotlib, seaborn)
• Hands‑on experience using UMAP, t‑SNE, PCA, or other representation analysis methods
• Experience analyzing data for both classical ML and deep learning models
• Strong understanding of ML fundamentals and model evaluation methodologies

Desirable

• Join the Evolv Machine Learning & Sensors team as a Data Scientist focused on driving deep understanding of sensor data, feature spaces, and data quality that power our AI/ML systems
• This hands-on role emphasizes representation analysis, exploratory data insights, and data-centric improvements that directly enhance model accuracy, robustness, and generalization
• You will work across classical ML and deep learning pipelines to identify blind spots, diagnose data issues, and guide data curation and collection strategies
• Success in the Role: What performance outcomes will you work toward in the first 6–12 months?
• In the first 30 days:
• Develop a strong understanding of Evolv’s sensor ecosystem, datasets, and ML pipelines
• Review dataset structure, labeling processes, and existing exploratory data analyses
• Run initial UMAP/PCA/t-SNE analyses to map data distributions and identify anomalies
• Identify opportunities to improve data quality, labeling consistency, and dataset coverage
• Within the first three months:
• Perform deep representation analysis across sensor, time‑series, and feature data
• Evaluate classical ML and deep learning models by linking model errors to data issues
• Define data quality metrics and initial dataset acceptance criteria
• Collaborate with data collection teams to guide targeted data acquisition and relabeling
• Data mining on existing field data and understanding patterns and extract useful information and insights
• Design methods to improve data quality, converting noisy/unverified data into clean/verified data
• By the end of the first year:
• Own data‑centric insights that directly improve ML model performance
• Establish ongoing monitoring of data drift, blind spots, and label quality
• Provide strategic guidance for future data collection, annotation, and curation
• Develop automated tools and dashboards for data quality reporting and representation analysis
• The Work: What type of work will you be doing? What assignments, requirements, or skills will you be performing on a regular basis?
• Data Understanding & Representation Analysis:
• Analyze high‑dimensional sensor and feature data using UMAP, t‑SNE, PCA, and related techniques
• Identify clusters, outliers, distribution gaps, and blind spots across classes and environments
• Diagnose dataset shift, domain mismatch, sparsity, and representation collapse
• Model‑Aware Data Analysis:
• Conduct data analysis aligned with both classical ML models (XGBoost, SVR, k‑NN, tree‑based models) and deep learning models (CNNs, Transformers)
• Analyze embeddings, confusion matrices, and failure cases to map model issues back to data causes
• Data Quality & Curation:
• Investigate imbalanced data, noisy sensor signals, and mislabeled or ambiguous samples
• Develop strategies for weakly labeled or unlabeled data using clustering or pseudo‑labeling
• Define data quality metrics, acceptance criteria, and labeling strategies
• Work with internal teams and external vendors to improve label consistency and coverage
• Insight‑Driven Improvements:
• Translate exploratory insights into clear recommendations for data collection, relabeling, or filtering
• Drive data‑centric improvements instead of relying solely on algorithmic changes
• Track KPIs such as data quality, data quantity, collection rate, and utilization efficiency
• Collaboration & Communication:
• Work closely with internal and external data collection teams to refine data pipelines
• Communicate findings through visualizations, reports, and technical deep‑dives
• Example Problems You Will Own:
• Use UMAP/PCA to detect under‑represented regions of sensor feature space and guide new data collection
• Identify mislabeled samples through clustering and model‑disagreement techniques
• Analyze class imbalance impacts across ML model families and propose data‑centric mitigations
• Cluster large volumes of unlabeled sensor data to support annotation prioritization
• What is leadership like for this role? What is the structure and culture of the team?
• You will join our R&D organization, reporting directly to VP of ML and sensors. In this role, you will interface with cross-disciplinary teams of highly skilled and autonomous engineers with expertise in Electromagnetics, Computer Vision, and AI. Our R&D organization includes more than 100 dedicated developers, engineers, scientists, managers and directors, each bringing deep technical knowledge and a strong culture of collaboration and support
• The team culture is one based on building trust, collaboration, on-going development through kindness, authenticity, courage, drive, and fun!
• Our software-enabled hardware products, this position requires a minimum of 60% or 3 days on-site work