Loon is building Project Bellwether and needs a Lead Data Scientist to help unify the data infrastructure behind multimodal model training pipelines. This onsite role in Mountain View, CA focuses on creating and scaling a multimodal geospatial data corpus, spanning automated ingestion, data quality validation, dataset versioning, and end-to-end development through production deployment for pre- and post-disaster response.
What you’ll do
- Architect and build automated ETL and ELT pipelines to aggregate, clean, and harmonize data from disparate sources, databases, and operational ingestion flows.
- Apply DataOps practices, including data quality monitoring, automated schema validation, and anomaly detection to identify corrupt or mislabeled data early.
- Standardize and integrate third-party annotation workflows and remote labeling feeds into unified datasets for model training.
- Design and maintain dataset versioning and storage systems to support reproducible machine learning experiments and reliable dataset retrieval.
- Collaborate with ML engineers and operations teams to transform geospatial, imagery, and sensor metadata into structured training features.
- Lead end-to-end development of multimodal machine learning models, from research and dataset architecture to production deployment.
- Define product and data strategy for the multimodal data corpus, leveraging proprietary datasets to support AI-driven disaster resilience tools.
- Work with engineering teams to optimize model performance, scalability, and integration into existing product workflows.
- Establish MLOps best practices, including model versioning, automated testing, and performance tracking for reliable, high-quality deployments.
- Help hire and grow the data science function by setting technical standards and shaping team culture as the group scales.
- Design, build, and maintain robust data ontologies and taxonomies to organize, categorize, and integrate complex geospatial and multimodal datasets.
Requirements
- Advanced Degree in Computer Science, Data Engineering, Software Engineering, or a related technical field.
- 7+ years of experience building scalable data pipelines, managing relational and non-relational databases, and unifying fragmented data storage systems.
- Expertise in Python and data manipulation libraries such as Pandas, NumPy, or SQL.
- Practical experience implementing automated data validation, quality control frameworks, and dataset versioning practices.
- Hands-on experience structuring datasets for machine learning workflows, including annotations, metadata tracking, and training set curation.
Technologies
- Python, Pandas, NumPy, SQL
- GDAL, GeoPandas, Rasterio
- Shapely, Zarr, COGs
It’d be great if you also had these
- Experience constructing automated ELT/ETL pipelines to aggregate, clean, and integrate complex geospatial, satellite image, and telemetry streams into unified repositories at scale.
- Ability to deploy comprehensive data validation systems, schema checks, and anomaly monitoring to support high-fidelity inputs for machine learning.
- Experience maintaining structured dataset versioning infrastructure for reproducible testing and efficient retrieval of operational features.
- Experience harmonizing external labeling pipelines and annotation feeds into streamlined, model-ready training assets.
- Experience partnering with software engineers, data scientists, and operational stakeholders to convert complex sensor attributes into targeted prediction features.
- Good knowledge of geospatial data sources and formats, plus familiarity with GIS tools and spatial analytics libraries such as GDAL, GeoPandas, Rasterio, Shapely, Zarr, and COGs.
Compensation: USD 207,000 - 311,000 per year.
Location: Mountain View, CA (onsite).