AI 先驱：将变革性的 GenAI 架构投入生产.pdf

《AI 先驱：将变革性的 GenAI 架构投入生产.pdf》由会员分享，可在线阅读，更多相关《AI 先驱：将变革性的 GenAI 架构投入生产.pdf（39页珍藏版）》请在三个皮匠报告上搜索。

1、 2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.S M B 3 0 1AI Pioneers:Shipping Transformative GenAI Architectures to ProductionPranav SharmaEMEA Lead SA-SMB,AWSHe/himJay RaoPrincipal Solutions Architect,AWSHe/him

2、Safwan AlmodayhanCEO,MISRAJ AIHe/him 2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.AgendaAI Pioneers OverviewAI Pioneer Architectures 2025 1.Building LLM Serving Platforms2.Creative Content Generation3.Misraj AI-Building Arabic Vision Language Model Emerging Architecture Pattern

3、 for 2026 and beyond 2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.AI PioneersBringing Innovative customer experiences to marketSmall and agile teams but have big AmbitionsDisrupting Industries 2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.2025,Amazon Web Se

4、rvices,Inc.or its affiliates.All rights reserved.AI Pioneer use case#1 Building LLM Serving Platforms 2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.What is an LLM Serving Platform?As LLM Inference requirements becomes omni-present in every application and at every layer,Organiza

5、tions have started offering LLM Serving capabilities at scale as a platform to applications and layers across the organization and to its customersLLM Inference is a fundamental building block of the modern tech stack 2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.LLM Serving Pla

6、tform Requirements Requirements Model ChoiceFrontier Models,OSS Models,Custom Models Supporting ServicesVector DB,Guard Rails,Security,Observability,.SaaS Capabilities Rate Limiting,Cost Attribution,Usage Reporting,.Self Managed ModelsSelf hosted models,Performance Optimizations,.Models AnywhereHost