Victor Rojo
Principal Global Tech Lead - Conversational AI Partners Competency at AWS
Seattle, Washington, United States
Actions
Victor Rojo has worked in the software industry for over two decades and is a huge fan of Conversational AI. In his current role as Principal Partner Solutions Architect at AWS, he is the driving force behind AWS's Conversational AI partners competency and is heavily involved in Generative AI initiatives, including those that deal with technologies like Amazon Lex, Amazon SageMaker Jumpstart, Amazon Bedrock, Amazon Transcribe, and LangChain. Before joining AWS, he was part of the team that launched the Amazon Alexa Skills Kit in Mexico and the US for Spanish-language speakers.
Links
Area of Expertise
Topics
Writing a Multi-LLM and Multi-RAG Powered Chatbot using AWS Gen AI
In this discussion, we will examine the AWS GenAI LLM Chatbot. Using this method, you may easily deploy and execute a wide variety of Foundation Models (FMs) in the cloud using Amazon Web Services (AWS).
It is compatible with numerous LLMs and Vector datastores for RAG. Amazon Bedrock (Amazon Titan, Anthropic's Claude, AI21 Jurassic, and Co:Here), Amazon SageMaker, and Amazon Aurora with PGVector, Amazon Open Search, and Amazon Kendra are all examples of these.
Define your assistant personality with Amazon Bedrock, Anthropic, and Gradio
In this show-and-tell session, you will learn how to leverage Amazon Bedrock and Gradio to define the attributes of generative AI assistants, so you don't start building and testing your prompts from scratch. A source code repository will be provided for experimentation.
Transforming the Conversational AI Lifecycle with Generative AI and LLMs
This is an exciting and forward-thinking session that delves into the transformative potential of Generative AI and Large Language Models in the field of Conversational AI.
Attendees will gain valuable knowledge and inspiration that will help them stay ahead of the curve in this rapidly evolving field.
Sample Outline:
1. The Expansion of task-oriented bots capabilities by integrating them with an LLM.
Examples of this:
A task-oriented bot can fallback to an LLM instead of answering “I don’t know that”
2. Leveraging LLMs to define and generate the following data in task-oriented bots during their implementation
-- Sample phrases for intents
-- Sample prompt variations as response to a user and its level of experience (first time users vs recurrent users)
-- Custom slot values and their synonyms to capture most of the different ways to match values
-- Custom vocabularies to increase the accuracy of Automatic Speech Recognition
-- Functional Test Data generation to validate both Happy Paths, Alternative Paths, and Exceptions (text and audio)
3. Leveraging LLMs for dynamic machine translation, documents processing, audio call transcriptions, when interacting with a user
4. Create sample code templates to handle intents or any other routine, including code templates for consuming APIs
5. Support for multimodal solutions that require use of multimedia content to improve the customer experience on the fly.
6. I planned to challenge the participants: Did you know that your LLM is has a gap of knowledge when you deploy it? Unless you train it frequently or if you use task-oriented bots that complement them (for example, to get the weather, money exchange rates, stock, traffic, news, etc.)
7. Explain how the top companies in cross industry will eventually become the creators of multiple XYZGPT LLMS? For example BloombergGPT, but why not think about others, like NetFlixGPT, BankingGPT, ShoppingGPT, TradingGPT, etc.
8. Creation of sample data
Victor Rojo
Principal Global Tech Lead - Conversational AI Partners Competency at AWS
Seattle, Washington, United States
Links
Actions
Please note that Sessionize is not responsible for the accuracy or validity of the data provided by speakers. If you suspect this profile to be fake or spam, please let us know.
Jump to top