BUILDING END-TO-END CONVERSATIONAL AI APPLICATIONS

This site requires Javascript in order to view all its content. Please enable Javascript in order to access all the functionality of this web site. Here are the instructions how to enable JavaScript in your web browser.

Workshop: February 18 2022, 2:00 to 7:15 p.m. IST

A PRACTICAL WALKTHROUGH WITH LAB SESSIONS

Join us for a workshop where we’ll take you through building and deploying SOTA conversational AI applications with real-world industry use cases.

Conversational AI solutions need to train complex models across multiple modalities. This requires a deployment solution that uses frameworks to capture and manage inputs from multiple sensors. The deployment needs to realize a micro services approach delivering easy integration and simple to use APIs. The catch, then, is to deliver this with high accuracy and real time response.

Over this day-long workshop, NVIDIA experts will take you through a detailed guided walkthrough starting with the state of conversational AI today, and then deep diving into the conversational building blocks.

We’ll also explore Industry use cases such as:

Smart video conferencing with real-time translation and text-to-speech with photo animation “live portrait” and eye contact
Enabling metaverse-ready AI avatars that are customizable and can communicate with customers, understanding their intent and then make meaningful recommendations
Creating a unique custom voice for conversational AI chatbot with 30 minutes of audio data for training
The NVIDIA Audio2Face Streaming Audio Player, which will enable a new generation of expressive facial animation from just an audio source powered by AI

Explore a hands-on walkthrough of the workflow involved in deploying of fine-tuned models and explore challenges related to performance optimization and scaling deployment. This will be followed by a session on efficiently developing and deploying Large-Scale Language Models like GPT3 to enable the build of domain-specific chatbots, personal assistants, and other AI applications that understand language with unprecedented subtlety and nuance.

SEE AGENDA

WORKSHOP DETAILS

Tools, libraries, and frameworks

The workshop will use end-to-end conversational AI tools from NVIDIA. Including Riva, NeMo^™, TAO Toolkit, TensorRT^™, and Triton^™.

Prerequisites

Basic Python programming experience
Fundamental understanding of a deep learning framework, such as TensorFlow, PyTorch, or Keras
Basic understanding of neural networks

Prerequisites for hands-on

This session can be attended in two modes-–with and without hands-on instruction.

If you're attending the hands-on session (which will include creating an application of conversational AI using NVIDIA Riva), please come prepared with a GPU-enabled system or cloud instance. Following are the prerequisites to try hands-on programming during the session:

You need to create an account on NVIDIA GPU Cloud (NGC) and have the access enabled.
For step-by-step instructions, refer to the NGC Getting Started Guide.
You’ll need a GPU system (on-prem or Cloud ) based on NVIDIA Volta, Turing™, or A100 GPU and have Docker installed. For more information, refer to the Support Matrix.

If you don’t want to or cannot come prepared with the prerequisites for hands-on instruction, you can attend the session and watch as the instructor demos how to build a conversational AI application.

AGENDA

Schedule	Topic	Instructor
02:00 pm to 02:30 pm IST	Intro (30 mins)
02:30 pm to 03:30 pm IST	Build a Conversational AI Application (60 mins) Conversational AI - State of Play Conversational AI Use Cases and Adoption Across Industries India Opportunity Building Blocks of Conversational AI Applications Challenges - Multiple Modalities, Complex Models, Sensor Fusion, Deployment while achieving High Accuracy with Low Latency Pre-Trained Models and Deployment at Scale (Technologies - NVIDIA Riva, NVIDIA TensorRT^™ & Triton^™) Fine tuning of Models and Custom Deployment (Technologies - NVIDIA TAO)	Meriem Bendris Senior Deep Learning Data Scientist
03:30 pm to 03:45 pm IST	Break (15 mins)
03:45 pm to 04:30 pm IST	Industry Use Cases - Demo (45 mins) Smart video conferencing Enabling AI Avatars ready for metaverse Create a unique custom voice for conversational AI chatbot Audio2Face Streaming Audio Player	Sunil Kumar Jang Bahadur Senior Solutions Architect Madhushree Basavarajaiah Solution Engineer
04:30 pm to 05:15 pm IST	Inference and Deployment (45 mins) Explore challenges related to performance, optimization, and scaling in production deployment of conversational AI applications Gain an understanding of the inference deployment process Analyze non-functional requirements and their implications Use a Helm Chart to deploy a conversational AI application with a Kubernetes cluster	Vikash Kumar Senior Alliance Manager, Developer Relations
05:15 pm to 06:15 pm IST	A Practical Walkthrough (60 mins) Handson for detailed workflow of building and deploying a voice based conversational AI chatbot for a domain-specific Q&A including ASR, NLU and TTS implementations Participants who come prepared with the GPU resources as given in the Prerequisites for hands-on will be able to do this part hands-on	Vikash Kumar Senior Alliance Manager, Developer Relations
06:15 pm to 06:30 pm IST	Break (15 mins)
06:30 pm to 07:00 pm IST	Efficiently Training Large-Scale Language Models (30 mins) Session will discuss how to efficiently develop and deploy Large-Scale Language Models like GPT3 to enable the build of domain-specific chatbots, personal assistants and other AI applications that understand language with unprecedented levels of subtlety and nuance. (Technologies - NVIDIA NeMo™, NVIDIA NeMo Megatron)	Meriem Bendris Senior Deep Learning Data Scientist
07:00 pm to 07:15 pm IST	Q&As (15 mins)