Join us for an enlightening conversation with Anton Alexander, AWS's Senior Specialist for Worldwide Foundation Models, as we delve into the complexities of training and scaling large foundation models. Anton brings his unique expertise from working with the world's top model builders, along with his fascinating journey from Trinidad and Tobago to becoming a leading AI infrastructure expert.

Discover practical insights on managing massive GPU clusters, optimizing distributed training, and handling the critical challenges of model development at scale. Learn about cutting-edge solutions in GPU failure detection, checkpointing strategies, and the evolution of inference workloads. Get an insider's perspective on emerging trends like GRPO, visual LLMs, and the future of AI model development.

Don't miss this technical deep dive where we explore real-world solutions for building and deploying foundational AI models, featuring discussions on everything from low-level infrastructure optimization to high-level AI development strategies.

Learn more: http://go.aws/47yubYq

Amazon SageMaker HyperPod : https://aws.amazon.com/fr/sagemaker/ai/hyperpod/

The Llama 3 Herd of Models paper : https://arxiv.org/abs/2407.21783

Chapters:

00:00:00 : Introduction and Guest Background

00:01:18 : Anton Journey from Caribbean to AI

00:05:52 : Mathematics in AI

00:07:20 : Large Model Training Challenges

00:09:54 : GPU failures : Lama Herd of models

00:13:40 : Grey failures

00:15:05 : Model training trends

00:17:40 : Managing Mixture of Experts Models

00:21:50 : Estimate how many GPUs you need.

00:25:12 : Monitoring loss function

00:27:08 : Crashing trainings

00:28:10 : SageMaker Hyperpod story

00:32:15 : How we automate managing grey failures

00:37:28 : which metrics to optimize for

00:40:23 : Checkpointing Strategies

00:44:48 : USE Utilization, Saturation, Errors

00:50:11 : SageMaker Hyperpod for Inferencing

00:54:58 : Resiliency in Training vs Inferencing workloads

00:56:44 : NVIDIA NeMo Ecosystem and Agents

00:59:49 : Future Trends in AI

01:03:17 : Closing Thoughts

Join us as we sit down with AWS Solutions Architect Mirabela Dan for a journey into the world of generative AI for developers. Whether you're a seasoned developer or just getting started, this episode is your gateway to staying ahead in the AI revolution.

Discover: the game-changing shift from "vibe coding" to spec-driven development and learn about the latest AWS AI tools that can supercharge your productivity. Get real-world insights on how AI is reshaping the developer landscape and plan essential strategies to future-proof your development career

Don't miss this power-packed episode where we demystify the intersection of AI and modern development practices. Get ready to transform the way you build with AWS's innovative AI solutions!

Learn more: https://aws.amazon.com/ai/ and https://kiro.dev/

Connect with Mirabela: https://www.linkedin.com/in/carmenmirabeladan/

Chapters

0:00:00 : Introduction and Guest Background

0:02:48 : Mirabela's Journey at AWS

0:05:38 : Working with Different Customer Types

0:08:52 : Shift from Infrastructure to AI

0:09:35 : Evolution of Developer Role with AI

0:15:49 : Challenges in Developing with AI

0:20:06 : AI for Code Generation vs Maintenance

0:28:33 : Spec-Driven Development Approach

0:39:30 : Balancing Planning vs Rapid Delivery

0:42:23 : Kiro - AWS AI-Powered IDE

0:44:46 : MCP Integration

0:46:10 : Agent Steering and Hooks

0:54:43 : Code Transformation with AI

0:55:57 : Future of Developer Role

0:59:55 : Consuming AI/Tech Updates

1:00:17 : Personal Learning Methods

1:01:30 : Optimism for Future of Development

1:02:34 : Closing Remarks

In this episode of AWS for AI, we sit down with Akshat Prakash, CTO and co-founder of Camb.ai, to explore how this Dubai-based startup is revolutionizing content localization through AI. From making sports accessible in 140+ languages to preserving indigenous cultures with fewer than 600 speakers, discover how they're breaking down global language barriers.

Witness history as we discuss how Camb.ai partnered with NASCAR to become the first company to livestream a race with real-time AI dubbing. Deep dive into their groundbreaking technical solutions - from preserving context, sarcasm, and emotion in speech, to their innovative approach of separating voice identity from speech prosody, solving critical ethical challenges in voice AI.

Get exclusive insights into their MARS and BOLI models, their partnership with AWS, and learn why deep problem understanding trumps technical expertise in building successful AI solutions. Whether you're a technologist, content creator, or business leader, this episode offers valuable insights into the future of global communication.

Learn more about CAMB.AI: https://www.camb.ai/

Chapters

0:00:00 : Dubbed Introduction

00:00:17 : Episode Introduction

00:02:14 : From Siri to Camb.ai: Akshat's AI Journey

00:03:52 : Breaking Language Barriers: The Camb.ai Family Story

00:05:47 : Beyond Translation: Understanding Cultural Context

00:07:00 : The Story Behind the Name 'CAMB'

00:08:01 : The Mamba Mentality

00:10:19 : Dubai to the World: Exporting AI Innovation

00:12:43 : Making History: First Multi-Language Live Race Stream

00:14:30 : A "Man on the Moon" Moment in AI

00:15:15 : Camb.ai Technology Suite and Offerings

00:19:56 : Creating Value Across All Layers

00:22:36 : Solving Last Mile Problems in AI Localization

00:23:59 : Focus on Results: Getting the Job Done

00:24:38 : Tackling the Hardest Challenge First: Live Sports

00:28:00 : MARS Architecture: Balancing Prosody, Speed, and Performance.

00:31:01 : Understanding Auto-Regression Tradeoffs

00:31:58 : Speaker Entanglement: Core Voice Identity Challenges

00:33:42 : Ethics in Voice Identity Usage

00:35:12 : Building with Resource Constraints

00:36:48 : The Case for Small Language Models

00:42:21 : Speech-to-Speech vs. Cascading Architecture Approach

00:46:30 : Preserving Context in Cascading Architecture

00:49:29 : BOLI: Enhanced Context Through Multi-modality

00:50:59 : Inclusive AI: Supporting All Languages

00:53:12 : Managing Dialects vs Languages

00:53:49 : MBC Partnership: Advancing Arabic Understanding

00:54:56 : From Sports to Rap: Diverse Use Cases

00:59:53 : A CTO's AWS Journey

01:02:38 : Accelerating Innovation with SageMaker Hyperpod

01:03:58 : The Future of AI: Final Thoughts

01:07:24 : Closing Remarks