Keep SageMaker. Add e6data underneath.
SageMaker notebooks, pipelines, and models read your open tables through e6data at up to 10x faster p95. Compute cost on those workloads drops by up to 60%. No migration, no rewrites, no second copy of your data.
One platform, two engines, each doing what it does best
Model training, tuning, and deployment, and the workflows your ML teams built.
Your data, in the same open Iceberg tables and catalogs you govern today.
Its role in your contracts and your roadmap. Nothing is ripped out.
High-concurrency reads from notebooks, models, and agents: 1,000+ QPS with p95 under 2 seconds.
Real-time work: streaming events queryable as open Iceberg in ~15 seconds.
Multi-region, multi-cloud, and hybrid reach, with ~99% less egress.
standard SQL over JDBC and the Python connector · same Iceberg tables · zero data movement
Feature engineering on open tables
Prepare training data with SQL directly on your lakehouse, at 1,000+ QPS when pipelines fan out, without exporting a second copy.
Real-time streaming to Iceberg
Kafka topics and CDC land as open Iceberg tables with sub-second ingest, queryable in ~15 seconds, ready for real-time features.
Multi-region, multi-cloud, hybrid
Query data where it lives, in any region, cloud, or on-prem environment, and cut egress by ~99% instead of centralizing everything first.
Native Iceberg querying
Read and write open Iceberg directly, with any catalog, so the lakehouse stays one copy of the truth instead of a second silo.
Don't take the word for it
The NASDAQ SaaS case ran 15M queries a day beside the platforms already in place, at p95 1.2s.
Amazon SageMaker compatibility questions
Does this replace Amazon SageMaker?+−
No. e6data runs alongside Amazon SageMaker on the same open tables and takes the workloads where it wins on cost or concurrency. Amazon SageMaker keeps everything else.
Do my queries need rewriting?+−
No. e6data speaks standard SQL over JDBC and the Python connector, so notebooks and pipelines query it directly.
What about governance and access control?+−
Your existing catalog policies keep enforcing table, column, and row-level access. e6data inherits them rather than duplicating them.
How do I know which workloads to move?+−
Start with the ones that dominate the bill or miss SLAs at peak. Bring one to a demo and we will run it side by side on your tables.
Book a demo on your own workloads
Reach out to book a demo, share challenges you're facing, and tell us how this fits into what you're currently working on or thinking about.
Prefer to self-serve? Problems we're solving →