GTC 2020: How SageMaker Helped NerdWallet Build an ML Platform
After clicking “Watch Now” you will be prompted to login or join.
Click “Watch Now” to login or join the NVIDIA Developer Program.
How SageMaker Helped NerdWallet Build an ML Platform
Sharadh Krishnamurthy , NerdWallet | Ryan Kirkman, NerdWallet
NerdWallet connects millions of users with the best financial products for them, and data science is an important piece. Six months ago, it took months for NerdWallet to launch and iterate on models—now, it takes days. Senior Engineering Manager Ryan Kirkman and Staff Software Engineer Sharadh Krishnamurthy will explain how NerdWallet massively increased its rate of data science learning by building a machine-learning platform on top of SageMaker and Airflow. They'll share how SageMaker's BYO container feature helped them design a human-centric data science workflow and reuse parts of their existing build/deploy system. They'll also share why building a machine-learning platform is now possible with a small team thanks to massive technological progress over the past 10 years, like GPUs that are effectively machine-learning supercomputers. What cost $100 million in 2008 can now be rented from Amazon Web Services for about $34 an hour (p3.16xlarge instance with 8 NVIDIA Tesla v100 GPUs).