TLDRocket
Sign in

Run Positron on Amazon SageMaker AI for data science workflows

Amazon Web Services Abhishek Maligehalli Shivalingaiah

Positron now runs on Amazon SageMaker AI. That lets data teams query, model, deploy, and report in one governed Space.

Based on reporting by Amazon Web Services, Abhishek Maligehalli Shivalingaiah — read the original for the full story.

Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error

AWS and Posit are trying to collapse a very familiar mess. Instead of bouncing between separate tools for data access, R, Python, deployment, app work, and reporting, data scientists can now run Positron inside an Amazon SageMaker AI Space and keep the whole workflow in one place.

The setup is not just a nicer editor bolted onto the side. Positron runs under the Space execution role, so it can query Amazon Athena, the AWS Glue Data Catalog, and Amazon S3 without storing keys. Compute comes from the SageMaker AI instance you launch, and teams can reserve capacity with training plans for scheduled work. If Posit Assistant is turned on, it can use Amazon Bedrock as its model provider, which keeps the AI help inside the customer’s AWS account and Region.

AWS walked through a recorded run using a synthetic portfolio of 50,000 loans. The source data lived in S3 and was registered in Glue. Athena found 1,500 records with missing income and 1,015 defaults, while the overall default rate came out to 2.03 percent. R handled feature work and data exploration, then Python trained an XGBoost classifier on 40,000 rows with three model features. The held-out AUC was 0.834, and the highest-risk decile showed a 12.3 percent observed default rate.

From there the workflow pushed scored results for 48,500 loans back into Athena, created a SageMaker AI model and real-time endpoint, and called it from a Shiny for Python app. Quarto then stitched the evidence together into a reproducible report. That’s the pitch here: one Space, one project, fewer handoffs. Not magical. Just less admin pain.

The details matter because AWS is also drawing a bright line around responsibilities. Posit provides the image and product support; the customer still owns identity, permissions, licensing, networking, logging, patching, and operational choices. That’s the real story behind the convenience: less glue work for the analyst, but no disappearing act for the platform team.

My take — AI-written commentary, not fact-checked reporting

This is the kind of product move that actually makes sense: keep the notebook, the model, the app, and the report in one place, then wrap it in AWS permissions instead of a pile of stray keys. The industry loves to sell AI as magic; boring integration is usually what ships. Also, nice to see “governed” used for something other than a deck slide.

Read more about this at: Amazon Web Services

Related stories

The daily briefing

Every AI story that matters, in your inbox by 8am.

TLDRocket reads all relevant sources, removes duplicate coverage, and summarises the day in two minutes. Follow companies and topics for alerts, or get the briefing in Slack. Free, no spam, unsubscribe anytime.