The performance symptom is classic Lambda cold-start variability. The API is used sporadically, so Lambda execution environments may be recycled between requests. The first request takes longer because Lambda must initialize an execution environment; subsequent identical requests are faster because the environment is already warm. Provisioned concurrency keeps execution environments initialized and ready to respond, producing more consistent latency. AWS Application Auto Scaling can manage provisioned concurrency with target tracking, which helps align provisioned capacity with usage patterns. DynamoDB on-demand mode already scales automatically for table throughput, and DAX would mainly help read-heavy caching scenarios, not Lambda initialization delay. CloudFront may reduce network latency for cacheable content, but it does not remove Lambda cold starts for dynamic API calls. (AWS Documentation)

===============