2025-07-30_Taipei_Summit_Startup_Day_Keynote.pdf

Transcript

1. © 2025, Amazon Web Services, Inc. or its affiliates. All rights reserved. T a i p e i | J u l y 2 9 - 3 0 , 2 0 2 5
2. © 2025, Amazon Web Services, Inc. or its affiliates. All rights reserved. Growing with GenAI Startups in Japan 針原 佳貴 S T A R T U P D A Y K E Y N O T E Yoshitaka Haribara, Ph.D. Sr. GenAI Startup Solutions Architect Amazon Web Services Japan
3. © 2025, Amazon Web Services, Inc. or its affiliates. All rights reserved. Yoshitaka Haribara, Ph.D. Sr. GenAI/Quantum Startup Solutions Architect, AWS Japan Visiting Associate Professor, QIQB, The University of Osaka Yoshitaka Haribara has over 6 years of experience as a Solutions Architect at AWS, where he works closely with Japanese generative AI startups, helping them build and scale their services on AWS globally. He is passionate about supporting Japanese model developers to build Japanese LLMs and deliver them on the Amazon Bedrock Marketplace. With a research background in harnessing quantum optical devices for hardware accelerated computing, Yoshitaka also guides Japanese customers in experimenting with quantum computing on AWS. He is committed to helping Japanese quantum hardware providers bring their technologies to the cloud through AWS. X @_hariby
4. © 2025, Amazon Web Services, Inc. or its affiliates. All rights reserved. AWS AI Stack S D K S F O R A G E N T S Vertically Integrated Flexible/OSS A M A Z O N B E D R O C K Models Capabilities AgentCore Knowledge Bases Amazon Nova 3P Models Optimization Guardrails Customization Runtime 1P Tools Gateway Identity Memory Observability I N T E R F A C E S & P R O T O C O L S ( M C P / A 2 A ) D A T A Nova Act Strands Agents A P P L I C A T I O N S Kiro Amazon Q AWS Transform Amazon Connect AWS Marketplace A I & A G E N T D E V E L O P M E N T S O F T W A R E & S E R V I C E S I N F R A S T R U C T U R E A M A Z O N S A G E M A K E R A I A I C O M P U T E AWS Trainium MLOps Model training Deployment Model building Fine-tuning Governance AWS Inferentia GPUs
5. © 2025, Amazon Web Services, Inc. or its affiliates. All rights reserved. 2 0 2 3 Se p . AWS LLM Development Support Program 2 0 2 4 F e b . METI GENIAC Start 2 0 2 4 J ul . & 2 0 2 5 AWS Japan GenAI Accelerator Program 2 0 2 4 Se p . & 2 0 2 5 AWS Generative AI Accelerator 2 0 2 4 Oc t. METI GENIAC Cycle2 AWS Generative AI Support Programs and METI GENIAC 2 0 2 5
6. © 2025, Amazon Web Services, Inc. or its affiliates. All rights reserved. Helped Japanese companies/organizations to build foundation models (FM) and large language models (LLM) AWS LLM Development Support Program (2023) Among the 15 selected companies, 12 chose EC2 Trn1 (AWS Trainium) instances for FM/LLM training. 10 companies performed Meta Llama 2-based (continued) pre-training. "While model architectures are rapidly evolving with new developments, companies should focus their development efforts on training strategies and data as core technical competencies." — Mr. Takeshi Suzuki, Ricoh "We conducted Llama 2 70B pre-training and fine-tuning using data that includes our proprietary collected customer support corpus. We achieved the highest performance among Japanese models in the Japanese MT-Bench." — Dr. Tomofumi Nakayama, Karakuri
7. © 2025, Amazon Web Services, Inc. or its affiliates. All rights reserved. ③ Developer Community ④ Go-To- Market Support ① Compute Resource ② Technical Engagement AWS Provided Support Through 4 Pillars for GENIAC Cycle2
8. © 2025, Amazon Web Services, Inc. or its affiliates. All rights reserved. Amazon Bedrock B R O A D E S T S E L E C T I O N O F F U L L Y M A N A G E D M O D E L S F R O M L E A D I N G A I C O M P A N I E S NOVA JAMBA CLAUDE COMMAND EMBED RERANK LLAMA RAY2 Highly efficient processing & grounded generation for long context lengths Professional-grade images with creative control, deployable at scale Advanced image & language reasoning Specialized expert models for agentic reasoning and multimodal tasks High-quality video generation with natural, coherent motion & ultra-realistic details Software engineering AI for large enterprises STABLE DIFFUSION MISTRAL MIXTRAL PIXTRAL MALIBU POINT Frontier intelligence & industry leading price performance Excels at complex reasoning, code generation, & instruction following Powering efficient, multilingual AI agents with advanced search & retrieval Advanced reasoning models that solve complex problems step-by-step DEEPSEEK-R1 CTRL + F for video data: unlock the full potential of enterprise video assets PALMYRA Purpose-built models for building & scaling AI agents across the enterprise MARENGO PEGASUS (Coming soon)
9. © 2025, Amazon Web Services, Inc. or its affiliates. All rights reserved. 100+ models from over 30 providers on Bedrock Marketplace EVOLUTIONARY SCALE WIDN CAMB.AI GRETEL ARCEE AI PREFERRED NETWORKS WRITER UPSTAGE NCSOFT STOCKMARK KARAKURI JOHN SNOW LABS LIQUID DATABRICKS CYBERAGENT HUGGING FACE STABILITY AI LG AI RESEARCH MISTRAL AI SNOWFLAKE NVIDIA
10. © 2025, Amazon Web Services, Inc. or its affiliates. All rights reserved. Sakana AI Tokyo-based unicorn startup, working on nature-inspired AI, started their first research “Evolutionary Model Merge” on AWS. 15 https://sakana.ai/evolutionary-model-merge/
11. © 2025, Amazon Web Services, Inc. or its affiliates. All rights reserved. Sakana AI Agentic AI system requires intelligent models. Sakana AI is leveraging foundation models on Amazon Bedrock for scientific discovery. https://sakana.ai/ai-scientist/
12. © 2025, Amazon Web Services, Inc. or its affiliates. All rights reserved. Sakana AI “By evolving different ways of merging the algorithms, we end up with a better merged model than a human could design by hand. Anytime you can get a computer to search through a space of solutions for you, you win. That beats a human trying do it manually because a computer can do it faster, try more things than you, and have more patience than you.” — Llion Jones, CTO “AWS deeply understands our workload and what we’re trying to achieve. They have helped us quickly overcome challenges, such as capacity issues. We are quite a small team, so we didn’t have a platform engineer to set up a cluster. It’s really easy to use AWS services which has made it simple to explore our research.” — Takuya Akiba, Research Scientist 17 https://aws.amazon.com/startups/learn/letting-nature-lead-how-sakana-ai-is-transforming-model-building
13. © 2025, Amazon Web Services, Inc. or its affiliates. All rights reserved. Thank you! Please complete the session survey in the mobile app 針原佳貴 Yoshitaka Haribara X: @_hariby