Case Studies

Challenge

Large-scale ad-serving infrastructure supporting NFL Sundays, NBA playoffs, WNBA, and peak Amazon sale events like Prime Day was provisioned for worst-case load and left running around the clock. The compute fleet was significantly over-provisioned during off-peak windows — overnight, weekends, between campaign bursts — resulting in millions of dollars in annual waste on idle capacity.

Solution

As a Systems Development Engineer at Amazon Ads, performed a comprehensive utilization audit across the compute fleet. Identified over-provisioned instances relative to actual CPU and memory usage across normal windows. Rightsized instances to match real workload demands, then designed and implemented demand-based auto scaling policies tied to ad-serving traffic patterns, campaign queues, and event-driven spikes. Introduced scheduled scaling for predictable cycles and target tracking policies for unpredictable load during major live events — so infrastructure scaled up for the Super Bowl and scaled back down after.

Results

• 30% reduction in annual AWS spend
• Compute costs aligned to actual demand instead of peak capacity
• Auto scaling handles campaign launch traffic spikes automatically
• Scheduled scaling eliminates overnight and weekend waste
• Zero degradation to ad-serving latency or analytics throughput

Challenge

A regional healthcare provider was running patient records, scheduling systems, and clinical applications on aging on-premises infrastructure. The hardware was approaching end-of-life, compliance audits were flagging gaps, and the IT team lacked the bandwidth to maintain both the legacy environment and meet expanding HIPAA requirements. They needed to move to the cloud without disrupting clinical operations or compromising patient data security.

Solution

Designed and executed a phased migration to AWS with HIPAA compliance built into every layer. Implemented a secured VPC architecture with private subnets for all PHI workloads, encrypted data in transit and at rest, and enforced strict network segmentation using security groups and NACLs. Configured VPN connectivity for secure access from clinical sites, deployed AWS PrivateLink to keep traffic off the public internet, and established comprehensive logging with CloudTrail and VPC Flow Logs for audit readiness. Executed the AWS Business Associate Agreement and developed operational runbooks for the internal team.

Results

• Full migration completed with zero unplanned clinical downtime
• HIPAA-compliant network architecture from day one
• All PHI isolated in private subnets with end-to-end encryption
• Audit-ready logging and monitoring across the environment
• Eliminated capital expense of on-premises hardware refresh

Challenge

A mid-market SaaS company was running their entire stack on self-managed EC2 instances — web servers, application servers, databases, and background job processors. The team spent significant time patching, scaling, and troubleshooting infrastructure instead of building product. Instances ran 24/7 regardless of actual traffic, and the operational burden was slowing down feature delivery.

Solution

Migrated workloads to fully managed services and serverless platforms that scale with demand. Replaced self-managed EC2 web and application servers with AWS Lambda and API Gateway for request-driven workloads. Moved from self-managed PostgreSQL on EC2 to Amazon RDS with automated backups and patching. Shifted background processing to Lambda functions triggered by SQS queues, and migrated static asset serving to CloudFront and S3. Each component now scales independently based on actual usage rather than provisioned capacity.

Results

• Significant reduction in monthly infrastructure costs
• Near-zero operational overhead — no more patching or capacity planning
• Infrastructure scales automatically with traffic and processing demand
• Engineering team shifted from infrastructure maintenance to product development
• Pay-per-use model eliminated cost of idle capacity during off-peak hours