Nubank's Storage & Databases engineering team turned a high-volume DynamoDB optimization opportunity into an autonomous, continuously running pipeline — executing 3,000+ remediations on PointFive's recommendations with zero ongoing engineer involvement and capturing 2.5x the savings potential of manual work.
Results at a Glance
- 3,000+ DynamoDB optimizations executed autonomously
- Zero engineer involvement after automation
- 2.5x more savings potential captured vs. manual remediation
- Dozens of optimizations applied automatically per day
Nubank
Nubank is one of the largest digital financial services platforms in the world, serving 131 million customers across Brazil, Mexico, and Colombia. As a cloud-native organization, Nubank relies heavily on AWS to support real-time financial products at scale — and DynamoDB plays a particularly critical role, powering the low-latency, customer-facing workloads that sit on the path of every transaction.
Nubank had already invested in cloud cost visibility and used existing tools for spend reporting and financial tracking. But those tools weren't designed to detect hidden waste inside DynamoDB configurations, or to identify optimization opportunities as usage patterns shifted over time. To close that gap, Nubank adopted PointFive.
A High-Volume Opportunity That Outpaced Manual Remediation
After integrating PointFive, a recurring optimization pattern surfaced: DynamoDB tables provisioned on the Standard table class when actual usage made Standard-Infrequent Access a more cost-effective fit. PointFive's Suboptimal Storage Type for DynamoDB opportunity type continuously flags these mismatches.
The pattern emerges because DynamoDB access patterns shift over time. Customer behavior evolves, new workloads come online, older ones fade — but AWS billing doesn't adjust automatically when usage patterns change. Without continuous monitoring, these mismatches accumulate quietly.
Over the first five months, an engineering taskforce addressed roughly 500 to 700 DynamoDB optimization opportunities per month using the AWS CLI. The work was meaningful — but the volume was the problem. At that pace, the team was capturing only a fraction of the total opportunity, with the rest competing for attention against other engineering priorities. Modeling the gap revealed something striking: the optimization potential the team was leaving on the table was more than 2.5x what manual remediation could realistically capture.
For an opportunity type this repetitive, manual remediation wasn't going to close the gap. Automation was the only path.
Validating the Risk Profile
Before automating, the Storage & Databases "Predict & Prevent" engineering team — whose mandate is optimizing the performance and cost of Nubank's database systems — needed to confirm that automated remediation was safe.
The team reviewed AWS documentation and ran internal experiments to assess the behavior of DynamoDB storage class changes. The conclusion was clean: switching between Standard and Standard-Infrequent Access is a cost-only change. No effect on performance. No data integrity risk. With the risk profile confirmed, the team had clearance to build.
Building the Pipeline
The Predict & Prevent team designed a custom automation pipeline that integrates PointFive's GraphQL API directly into Nubank's internal automation infrastructure. The pipeline continuously evaluates PointFive's recommendations and applies remediations without human intervention.
Critically, the pipeline supports bidirectional optimization. When access patterns decline, tables move from Standard to Standard-Infrequent Access. When usage rises again, the pipeline reverses the change. DynamoDB usage isn't static, and a one-time migration wouldn't stay optimized as patterns shift. PointFive continuously evaluates usage patterns and surfaces updated recommendations; the pipeline consumes them and applies the appropriate change on an ongoing basis.
The initial build required upfront engineering investment — but the team built it as a reusable framework. The same pipeline architecture can be extended to support additional PointFive opportunity types as they prove suitable for automation.
Rollout was deliberately incremental. Nubank introduced automated remediation gradually to validate consistency and confirm reproducible outcomes. Once the team had confidence in the results, they scaled execution to dozens of automated optimizations per day — the current operating pace.
Results
Since automating DynamoDB storage class optimization on PointFive's recommendations, Nubank has executed more than 3,000 remediations.
Because the pipeline runs continuously, new opportunities are detected and resolved in real time. Engineers are no longer involved in routine DynamoDB tuning. Table owners are typically unaware that optimizations are occurring — the highest bar an efficiency program can clear.
What began as a targeted DynamoDB initiative has become a repeatable framework. Nubank plans to extend autonomous remediation to additional PointFive opportunity types and expand the pipeline across more regions and countries.
For Nubank, autonomous DynamoDB optimization isn't just an efficiency win — it's a model for how cloud cost management should work at scale: detection that's continuous, remediation that runs itself, and engineers who get to focus on shipping.
About PointFive
PointFive is the Cloud & AI Efficiency Engine. By combining a real-time cloud and infrastructure data fabric with AI-driven detection and guided remediation, PointFive transforms efficiency from a reporting exercise into an operational discipline. Customers achieve sustained improvements in cost, performance, reliability, and engineering accountability, at scale.
To learn more, book a demo.
