AWS re:Think Podcast

Companies now have a wide array of choices when it comes to Large Language Models. In this episode we talk to Ivan Lee, CEO of datasaur.ai about choosing the right LLM for your AI projects. We discuss open source vs proprietary models and the metrics for evaluating these models. We also cover Small Language Models as well and when to use more specialized vs general purpose models.

Datasaur:

https://datasaur.ai/

LLM Labs from Datasaur:

https://datasaur.ai/llm/home

AWS Hosts: Nolan Chen & Malini Chatterjee

Email Your Feedback: [email protected]

Companies need to provide real time insights to both customers and internal users. These insights power use cases such as personalization and fraud detection. StarTree Cloud is a real-time analytics platform built on Apache Pinot for building such applications that depend on real time insights. In this episode we meet with Chinmay Soman, Head of Product at Startree.ai to discuss the different dimensions of real-time analytics and how Apache Pinot and StarTree Cloud offer a robust platform for providing such insights to applications.

AWS Hosts: Nolan Chen & Malini Chatterjee

Email Your Feedback: [email protected]

Resources:

StarTree:

https://startree.ai

StarTree community Slack:

https://communityinviter.com/apps/startreedata/startree-community

Apache Pinot Slack: 

https://communityinviter.com/apps/apache-pinot/apache-pinot

Serverless / Free forever workspace:

https://stree.ai/free

Today we join Maria Khalusova, Staff Developer Advocate with Unstructured.IO, to discuss how companies can unlock their Unstructured Data to deliver better results from their Large Language Models. We talk about how Unstructured Data can enhance the performance of RAG applications, RAG vs Fine Tuning, data Chunking, Multi-Modal models and more. 

AWS Hosts: Nolan Chen & Malini Chatterjee

Unstructured Enterprise Platform beta signup: 

https://unstructured.io/platform

Embedding models MTEB Leaderboard: 

https://huggingface.co/spaces/mteb/leaderboard

2019 Deloitte report (source of the statistics that only 18% of organizations were using unstructured data):

https://www2.deloitte.com/us/en/insights/topics/analytics/insight-driven-organization.html

80% of data is unstructured, source: https://mitsloan.mit.edu/ideas-made-to-matter/tapping-power-unstructured-data

Papers showing RAG outperforming fine-tuning: 

https://arxiv.org/abs/2312.05934

https://arxiv.org/abs/2401.08406

Email Your Feedback: [email protected]

In this podcast episode, we explore with Dr. Alok Aggarwal, CEO and Chief Data Scientist at Scry AI. about the past, present, and future of artificial intelligence. He traces the evolution of AI from its early beginnings in the 1950s to its current state, marked by rapid advancements and the emergence of the fourth industrial revolution. We listen to the story behind Scry AI. and how we explore the world of generative AI.

If you are interested to learn more about his book "The Fourth Industrial Revolution & 100 Years of AI (1950-2050)”, here is the link to purchase it from Amazon: https://a.co/d/arodd1H

AWS Hosts: Nolan Chen & Malini Chatterjee

Email Your Feedback: [email protected]

In this podcast episode, we delve into the crucial topic of security related to AI and Machine Learning, with a particular focus on Generative AI. As AI and ML technologies rapidly advance, it is imperative to implement a robust security strategy based on the principle of Defense in Depth. We'll explore the potential security and privacy risks associated with AI/ML systems, highlighting the importance of Responsible AI practices. Additionally, we'll discuss practical approaches to implementing security measures within specific AI/ML services, such as Amazon Bedrock, a secure and compliant foundation for building and deploying AI/ML applications on AWS. Join us as we navigate the intricate landscape of AI/ML security, equipping you with the knowledge and best practices to safeguard your AI/ML deployments and mitigate potential risks.

AWS re:Inforce 2024:

https://hub.reinforce.awsevents.com/attendee-portal/catalog/?filters=5B4D572F-79F0-EE11-81DE-9CF328ECC866&search=APS351

AWS Hosts: Nolan Chen & Malini Chatterjee

Email Your Feedback: [email protected]

The amount of streaming data for Observability continues to grow exponentially from GBs to TBs daily. Join Edge Delta’s CEO Ozan Unlu as we discuss how their AI Observability Platform can deliver more Sustainable Observability, and why the needs of different size data volumes from different industries requires us to rethink Observability architecture in the Cloud. Ozan also offers valuable career advice for fast changing times.

AWS Hosts: Nolan Chen & Malini Chatterjee

Email Your Feedback: [email protected]

Many companies are sampling away data and collecting only trends. Join Axiom’s Chief Extrovert Kendall Miller as he explains why sampling your data can disrespect your individual users. 

AWS Hosts: Nolan Chen & Malini Chatterjee

Kendall Miller: https://www.linkedin.com/in/kendallamiller/

Email Your Feedback: [email protected]

In this episode we are joined by one of our AWS customers, Mixbook with DJ Charles, CTO and David Newhoff, CPO to share how they are leveraging generative AI to empower creativity and deepen human connection by unlocking the future of storytelling. Join us to hear how Mixbook took an idea from concept to production to empower their end-user’s creativity with AI-driven recommendations for images, backgrounds, stickers, captions, and more.

Try out the app: https://www.mixbook.com/

AWS Hosts: Nolan Chen & Malini Chatterjee

Email Your Feedback: [email protected]

In this episode, we join AWS Solutions Architects, Krishna Adettiwar and Eric Kiesow, to discuss how leveraging a multi-account strategy can help you protect your workloads, increase pace of innovation, and give you more cost visibility.

AWS Hosts: Nolan Chen & Malini Chatterjee

Email Your Feedback: [email protected]

Visit these resources to get started with the multi-account strategy:

• Organizing Your AWS Environment Using Multiple Accounts (AWS Whitepaper) - https://docs.aws.amazon.com/whitepapers/latest/organizing-your-aws-environment/organizing-your-aws-environment.html
• Multi-account strategy for small and medium businesses (AWS blog) - https://aws.amazon.com/blogs/mt/multi-account-strategy-for-small-and-medium-businesses/
• AWS Cloud Operations & Migrations Blog - https://aws.amazon.com/blogs/mt/

In this episode we meet with Matthew McClean, a Sr Manager from AWS's Annapurna Team to talk about Accelerators - the chips that make AI possible. We cover different Accelerators - GPUs, Trainium, Inferentia, Graviton and more.

AWS Hosts: Nolan Chen & Malini Chatterjee

Email Your Feedback: [email protected]

Links for the Show:

AWS re:Invent 2023 - Behind-the-scenes look at generative AI infrastructure at Amazon (CMP206)

https://www.youtube.com/watch?v=fDk09hms8s8

AWS re:Invent 2023 - Behind-the-scenes look at generative AI infrastructure at Amazon (CMP206)

https://www.youtube.com/watch?v=VPvguzeWlbU

Welcome to AWS Neuron

https://awsdocs-neuron.readthedocs-hosted.com/en/latest/index.html

Optimum Neuron

https://huggingface.co/docs/optimum-neuron/index