2026年2月

Kubernetes × AI × 大模型推理,一场社区共创的技术盛会

Kubernetes 成为 AI 基础设施的事实标准,

大模型推理进入工程化与规模化阶段,

云原生与 AI,正在真正走向融合。

于是,KCD Beijing 与 vLLM 社区 决定一起做一件事:

👉 把 Kubernetes 社区 与 大模型推理社区,带到同一个舞台。

KCD Beijing + vLLM 2026, 正式启航 🚀

🌏 关于 KCD Beijing + vLLM 2026

Kubernetes Community Days(KCD) 是 CNCF 指导的、由社区发起并组织的 Kubernetes 技术大会,强调:

• 社区驱动

• 真实实践

• 工程经验分享

KCD Beijing + vLLM 2026 将由 KCD Beijing 社区 与 vLLM 社区联合主办,这是一次:

云原生社区 × 大模型推理社区的深度联合共建

我们希望它不仅是一场会议,更是一场围绕以下主题的技术连接:

• Kubernetes 在生产环境中的真实经验

• AI / ML Infra 的系统性思考以及 AI 技术相关的实践

• vLLM 在大模型推理中的工程实践

• 从集群调度到模型服务的完整链路

🧭 活动基本信息

• 📛活动名称: KCD Beijing + vLLM 2026

• 📅 活动时间: 2026 年 3 月 21 日

• 🏗 联合主办方:

• Kubernetes Community Days Beijing

• vLLM Community


🎯 三大分会场 · 覆盖完整技术栈

本次大会将设置 三个分会场,从基础设施到模型推理,形成完整闭环:

Kubernetes 专场

• Kubernetes 生产实践

• 平台工程 / 多集群治理

• 网络、存储、安全、调度

• CNCF 生态项目经验

🤖AI / ML 专场

• AI / ML Infra 架构设计

• GPU / 异构算力调度

• 训练与推理平台建设

• AI 相关技术的实践

• AI 与云原生的结合方式

🚀vLLM 专场

• vLLM 架构与核心原理

• 大模型高性能推理实践

• vLLM + Kubernetes 的落地案例

• 推理性能优化与资源管理

• vLLM 生态与未来方向

如果你在 Kubernetes 上运行 AI 工作负载,
如果你在 使用 vLLM 构建推理服务,
如果你在 研究和使用 AI 相关的技术,

那么这里,一定有你的舞台。

🎤 Call for Proposals(议题征集)

我们诚挚邀请 工程师、架构师、维护者、社区贡献者 投稿分享:

演讲形式

• 标准演讲(Presentation)
⏱ 30 分钟 · 单人 / 双人

• 闪电演讲(Lightning Talk)
⚡ 10 分钟 · 快速、直接、聚焦一个点

我们更看重:

• 真实经验

• 实际问题

• 清晰的思考

而不是“完美故事”。

🗓重要时间节点

• 📢CFP开启: 2025 年 12 月 5 日

• ⏰CFP 截止: 2026 年 2 月 24 日

• 📩录用通知: CFP 截止后 1–2 周内

📝如何提交议题?

1⃣ 准备一份清晰的摘要

• 你要讲什么?

• 适合谁听?

• 听众能收获什么?

2⃣ 选择分会场与演讲形式

3⃣ 提交你的提案 👉
https\://sessionize.com/kcd-beijing-2026/

🤝感谢赞助商与合作伙伴

KCD Beijing + vLLM 2026 的举办,离不开赞助商和社区伙伴的支持。

特别感谢 (排名不分先后)

• AWS

• Red Hat

• 华为

• 蚂蚁开源

• OceanBase

• KubeEvents

• AtomGit

你们对 云原生与 AI 开源生态 的持续投入,让社区能够走得更远。

📌关于赞助与合作伙伴的详细介绍,将在下一篇文章中发布,敬请期待。

🌱为什么你应该参与?

• 这是 KCD Beijing × vLLM 社区的联合主办

• 这是一次 Kubernetes 与大模型推理的正面交汇

• 这是一场 以工程实践为核心的社区大会

• 这是一个 让一线工程师声音被听见的舞台

无论你是:

• Kubernetes / 平台工程师

• AI / ML Infra 开发者

• vLLM 使用者或贡献者

• 云原生或 AI 社区长期参与者

KCD Beijing + vLLM 2026,都欢迎你。

报名参会:

https\://www.bagevent.com/event/kcd-beijing-2026

🙌 最后

🎯 议题征集中

🎤 讲者招募中

🌱 社区共建中

欢迎转发给你的同事和朋友,也欢迎你,亲自站上这个舞台。

2026 年 3 月 21 日,北京见。

阅读更多 Voice Agent 学习笔记:了解最懂 AI 语音的头脑都在思考什么

前两天开了一个 Max20 账号,只用了半天就被风控的。今天就一直查找教程,看看有没有解决办法。然后搜索着就看到这篇文章:

https://www.augmunt.com/blog/claude-account-ban-solutions-deep-dive-2026/

涵盖 TLS 指纹 (JA4+)、TCP/IP 协议栈伪装、Canvas 熵值分析、行为生物统计学等硬核技术。真的有这么多风控点招呼上普通用户了么,那不是开账号以后很难。

如果你也觉得现在的 AI 不该只会打字——而是应该能开口说话、边听边说、自然接话/被打断,甚至还能自己跑流程——那 3 月 1 日在日本线下举办的 VOX TOKYO 就是一次很适合“直接上手”的线下冲刺。

这是一场聚焦社会课题解决的线下黑客松。我们邀请开发者、创业者和创意人才齐聚川崎,利用 TRAE 的高效开发能力、MiniMax 的自然语音合成,以及 Agora 的实时交互技术,在 8 小时内做出一个能 Demo 的语音 AI Agent 原型。新手也欢迎,现场有人一起 vibe coding,餐饮饮料都安排好,你只要带电脑 + 一个点子来就行。

作为 RTE 开发者社区关注的实时交互(Real-Time Engagement)场景,我们也会鼓励大家用 TEN Framework 快速搭起实时语音链路:把“听—说—对话节奏/打断—工具调用”的关键环节串起来,让你把精力花在 Agent 的核心逻辑和产品体验上,而不是被底层实时音频、延迟和稳定性拖住。

挑战方向也很“有戏”:👵银发陪伴与记忆照护、🗼访日游客沉浸式文化导览、⚙中小企业运营自动化(预约/库存等)、👨🏻🏫24/7 个性化学习导师。最后会有 Demo 展示与评奖(奖金池 10 万日元)以及交流环节——很适合把“语音 Agent”从概念推进到一个能拿出来讲的作品。

报名链接:
https://luma.com/j1pfdg53

Vox Tokyo: A Voice AI Hackathon for Social Good

✨ About the Event

Building with Voice AI used to be reserved for specialists. Not anymore. With the rapid rise of LLMs and AI agents, the barrier to entry has never been lower — and the possibilities have never been bigger.

Vox Tokyo is a mission-driven hackathon where ideas and experimentation come first.

  • Beginners welcome: New to AI development? Jump right in — this is your chance to get started.
  • Learn from the best: Industry pioneers and vibe-coding pros will be in the room and ready to help.
  • Just show up: Food and drinks are on us. All you need to bring is an idea.

You don't need to build anything perfect. Come explore what's possible when you put "voice" at the center of solving real problems.

🍽 Catering and drinks provided throughout the day.

📍Event Details

  • 📅 Date: Sunday, March 1, 2026
  • 📍 Venue: Fujitsu Uvance Kawasaki Tower (1-5 Omiya-cho, Saiwai-ku, Kawasaki, Kanagawa — JR Kawasaki Tower)
  • 👥 Format: In-person / Single-day event
  • 🧑💻 Tech Support: Hands-on guidance for vibe-coding and Voice AI models

🧩 The Challenge: Solve Social Issues with Voice AI

Participants will have 8 hours to build autonomous voice agents across four critical focus areas:

  • 👵 Silver Voice: A compassionate AI companion designed to combat elderly loneliness and support memory care.
  • 🗼 Omotenashi AI: An immersive "insider" guide that decodes Japanese culture and hidden gems for international visitors.
  • ⚙ Agentic Workforce: An "invisible operations manager" automating tedious tasks for small businesses facing labor shortages.
  • 👨🏻🏫 Agentic Education: A hyper-personalized AI mentor that provides 24/7 adaptive tutoring and roleplay practice.

Welcome to Vox Tokyo. Let's listen to the voices that matter — and build the technology to answer them. 🚀

👥 Participation

Application Deadline: February 27, 2025 (Fri) 13:00

How to Join:

  • In-person only
  • No development experience required — beginners welcome
  • Individual participation, or online team formation (up to 2 members)
  • Please bring your own laptop.

Eligibility: TRAE users (Sign up and download TRAE here)

Application:

  • Spots are limited.
  • Accepted applicants will be notified by email.

🏆 Prizes

💰 Total Prize Pool: ¥100,000 JPY

  • Grand Prize: ¥50,000
  • Runner-Up: ¥30,000
  • Third Place: ¥20,000
  • Special Award (3 teams): TRAE original merchandise

※ All prize amounts are pre-tax.

⏱ Event Schedule

12:30 - 12:45 Doors Open & Check-in

12:45 - 13:00 Opening Remarks

13:00 - 13:30 Sponsor Tech Sessions

13:30 - 18:00 Hacking Time (Main Session)

18:00 - 18:30 Project Submission & Final Tweaks

18:30 - 19:30 Final Presentations & Demos

19:30 - 20:00 Networking & Refreshments (Food & drinks provided)

20:00 - 20:30 Awards Ceremony & Closing

20:30 Event Ends

■Host

TRAE

TRAE (The Real AI Engineer) is ByteDance's AI-powered IDE that transforms how software is built. With intelligent IDE Mode for assisted coding and groundbreaking SOLO Mode for end-to-end automation, TRAE turns natural language into complete, functional applications.

We believe in rapid innovation and empowering builders of all skill levels. Our mission is simple: eliminate the gap between ideation and implementation—so you can focus on creating something extraordinary.

TRAE — Where Human Creativity Meets AI Intelligence.

MiniMax

MiniMax is a global AI foundation model company founded in 2021, committed to advancing the frontiers of artificial intelligence towards AGI. The company develops cutting-edge large language and multimodal foundation models, powering innovative products across video, image, speech, and music generation.

MiniMax's flagship products include Hailuo AI for video generation and MiniMax Audio for natural text-to-speech and voice cloning. With breakthrough innovations like Lightning Attention enabling 1-million-token context windows, MiniMax continues to push AI boundaries.

Together, we are building AGI with everyone.

Agora

Agora powers real-time voice, video, and interactive experiences at scale, trusted by thousands of developers worldwide to seamlessly embed communication into any application.

Voice AI is reshaping how we interact with technology. With Agora's SDKs and the TEN (Transformative Engagement Network) framework — an open-source voice AI toolkit for rapid prototyping — you'll have everything you need to build next-generation voice experiences. TEN's modular components for STT, LLM, and TTS integration make it ideal for hackathon innovation.

Join us and build the future of voice!

CreatorLabo

CreatorLabo is a Tokyo-based co-creation community of entrepreneurs, AI developers, and creators bridging global AI innovations with Japan's startup ecosystem.

We run hackathons like "Minna Hackathon," partner with global AI companies for localized developer education, and connect early-stage founders with talent to launch MVPs — all built around rapid prototyping and real-world AI application.

Our mission is developing Japan's next generation of entrepreneurial leaders with the tools, community, and global connections to validate, build, and create impact.

■ Co-host

Tokyo Design

Japan’s ultimate product design community for hands-on learning, AI-native practice, and real-world collaboration.

Voice AI / Space

Ahoy! Voice AI Space is your beacon to master voice tech's wild seas. We guide developers, entrepreneurs, and enthusiasts to top tools, news, knowledge, and careers, empowering everyone in the vast voice AI ocean.

Community Partner

MeltiingHack

MeltingHack is a vibrant, Tokyo-based hackathon community that brings together a diverse melting pot of developers from across Japan and around the world.

WaytoAGI

WayToAGI is one of the world’s largest open-source AI knowledge communities, dedicated to advancing the development of Artificial General Intelligence (AGI) and promoting the adoption and practical use of AI technologies worldwide.

Founder Institute Japan

Founder Institute (FI) is a leading startup accelerator helping founders turn ideas into fundable startups and grow globally. Since 2009, FI has guided early-stage founders through FI Core (idea to pre-seed): https\://fi.co/program

Alumni get ongoing access to Founder Lab: https\://fi.co/scale and can connect with investors via the FI Venture Network: https\://fi.co/investors

FI alumni have raised $1.5B+ across 200+ cities on 6 continents.

■ Sponsor

Shisa.AI

Shisa.ai is a Silicon Valley-led startup revolutionizing Japanese-specific AI through cutting-edge Open Source LLM development. We bridge world-class technical expertise with deep local insights to drive the next wave of AI innovation from Japan to the world.

Zilliz

Zilliz is the creator of the Milvus open-source vector database and a leader in high-performance vector database technologies for AI applications. Its infrastructure empowers organizations worldwide to unlock the value of unstructured data and accelerate AI development.

Supabase

Supabase is a Postgres development platform that gives you a production-ready backend out of the box: a managed Postgres database, Auth, Storage, instant APIs, Edge Functions, and Realtime subscriptions—plus support for storing vector embeddings for AI apps.

阅读更多 Voice Agent 学习笔记:了解最懂 AI 语音的头脑都在思考什么