Data Scientist – Geospatial Foundation Models

About SatSure

SatSure is a deep tech, decision intelligence company working at the nexus of agriculture, infrastructure, and climate action — creating impact for the other millions, with a focus on the developing world. As part of this mission, we're building geospatial foundation models that learn directly from Earth observation data — optical, SAR, and elevation — at scale. This role sits at the heart of that effort: architecting and training large-scale models that can generalize across geographies, sensors, and time. You'll be shaping the core intelligence layer that powers insights for millions, not just fine-tuning someone else's model.

Role

In foundation model development, data is the moat . You will drive the transformation of petabytes of raw geospatial data into a high-quality, high-entropy training and evaluation corpus .

This role sits at the intersection of remote sensing, data engineering, and ML , ensuring that models learn from diverse, representative, and well-curated data at scale .

Key Responsibilities

Data Curation & Pre-training Datasets

• •Design and implement data curation pipelines for large-scale pre-training datasets
• •Develop sampling strategies to ensure:
• •Geographic and biome diversity
• •Coverage across seasons, sensors, and resolutions

• •Mitigate dataset biases (e.g., over-representation of cloud-free or high-income regions)
• •Balance trade-offs between data quality, diversity, and scale

Evaluation Frameworks (Earth-Bench)

• •Design and own a comprehensive evaluation framework (“ Earth-Bench ”) to assess:
• •Representation quality (post-SSL embeddings)
• •Transfer performance on downstream tasks:
• •Segmentation
• •Yield prediction
• •Disaster mapping

• •Define metrics and benchmarks that reflect real-world generalization across geographies and time
• •Continuously evolve evaluation as new datasets, sensors, and tasks emerge

Data Systems & Pipeline Thinking

• •Build and maintain scalable data pipelines for ingestion, processing, versioning, and access
• •Work with ML and platform teams to:
• •Enable efficient data loading and training at scale
• •Optimize storage formats and access patterns (e.g., chunking, caching)

• •Ensure datasets are:
• •Reproducible
• •Well-documented
• •Easily usable across teams

Data-Centric ML Thinking

• •Analyze how data quality, diversity, and freshness impact model performance
• •Partner with researchers to:
• •Identify failure modes driven by data gaps
• •Improve datasets to unlock model gains (not just model changes)

Treat data as a first-class lever for improving model quality

Preferred Background

Domain Expertise

• •3–5 years of experience in Applied Data Science at scale
• •Strong understanding of remote sensing fundamentals , including:
• •Atmospheric correction
• •SAR backscatter
• •Orthorectification

Familiarity with multi-sensor data (optical, SAR, DEM, etc.)

Data Engineering at Scale

• •Experience working with large-scale (TB–PB) datasets across the ML lifecycle
• •Hands-on experience with:
• •Distributed data processing
• •Efficient storage and retrieval strategies

Understanding of how data pipelines interact with model training workflows

Tooling (Geo Stack)

• •Experience with geospatial data tooling, such as:
• •Xarray, Dask, Rasterio, Zarr
• •Google Earth Engine (nice to have)

Mindset

• •Strong data intuition —ability to reason about bias, coverage, and representativeness
• •Systems thinking: understands how data decisions impact model behavior at scale
• •Comfortable working in ambiguous, evolving problem spaces

Benefits:

• •Medical Health Cover for you and your family, including unlimited online doctor consultations
• •Access to mental health experts for you and your family
• •Dedicated allowances for learning and skill development
• •Comprehensive leave policy with casual leaves, paid leaves, marriage leaves, and bereavement leaves

Interview Process:

• •Intro call
• •Assessment
• •Presentation
• •Interview rounds (ideally up to 3-4 rounds)
• •Culture Round / HR round