How we are building serverless architectures for Deep Learning & NLP at Episource: Manas Ranjan Kar

375 views · Published 4 August 2017 · 19:51 · Indexed 4 November 2026

Channel: Hasgeek TV · 2017 · Science & Technology

Watch on YouTube

Serverless is the new kid on the block, and an exciting one at that ! As Anand Chitipothu puts it, it’s rapidly becoming the Uber of cloud computing resources.

At Episource, we have been working on creating a scalable NLP pipeline for scalable information extraction from medical discharge summaries. However, processing millions of charts can be expensive too. In this talk, we will show how we created a serverless and event driven architecture for deep learning. The platform requires no manual intervention and minimal upkeep & maintenance. We will share the challenges in creating such an architecture, especially when requirements of immutable configuration and heavy data workloads don’t go away easily. A detailed demo of the architecture would be showcased as well.

The audience can expect to learn a lot from our experiences and maybe go serverless at their roles as well !

Slides and more info: https://fifthelephant.talkfunnel.com/2017/44-how-we-are-building-serverless-architectures-for-d

More from this channel