Data strategies to enable ML and analytics at scale
summary
Canva shifts from daily batch exports to a streaming pipeline—CDC to Kinesis to S3 pipeline—so changes reach the warehouse in minutes; data platform handles hundreds of billions of events per week and 15 petabytes doubling every six months. Models run on SageMaker and EKS, Redshift ML trains in-SQL, DynamoDB serves a perceptual-hash index of over 10 billion images at ~2,000 qps, and productised features like Text-to-Image deliver scale—40 million images generated—proving the approach works.
AI-written summary · about
chapters (10)
- 0:00Intro
- 2:06Artificial intelligence maturity
- 4:02Value driven data roadmap
- 10:48Low code / no-code tools
- 15:56Canva data platform scale
- 17:07Foundations that scale
- 19:04Value driven product roadmap
- 20:43Experiments-drive innovation velocity
- 21:48Standardisation - Integration across Canva
- 22:33Democratisation-Service aligned data platform
related talks
29mData foundation for analytics and AI
29mBecoming a data-driven enterprise
29mSimplifying data flows
21mAWS On Air ft. Amazon SageMaker Canvas
44mBest practices for analytics and generative AI on AWS(ANT302)
19mDrive transformation through machine learning
43mData warehousing reinvented for today's needs
52mPreparing data for reporting, ML, and AI on AWS