Talk 1: Determined: open-source deep learning training platform
While deep learning (DL) has enormous potential, building DL-powered applications remains difficult, expensive, and time-consuming for most companies. A major cause is that deep learning engineers are forced to spend most of their time on DevOps and writing boilerplate code for common tasks like multi-GPU training and fault tolerance, rather than building better models.
Talk 2: Migrating from Hive to Open Source Delta Lake+Hive
In this talk, we introduce the story of adopting open source Delta Lake in our hybrid cloud environment and challenges we encounter along with the solutions we have implemented. The topic includes table conversion performance (6 hours to 9 minutes), cross cluster Delta Lake management (on-premise and cloud), transaction log updates via REST API and roll back functionality.
Yishuang Lu, works as a senior software engineer at Amobee. Yishuang is a graduate student at Carnegie Mellon University