James Liu

Taiwan Electronic Health… • 3K followers

《 中山附醫 × Azure OpenAI｜AI 民主化醫療新典範 》恭喜 中山醫學大學附設醫院 以 Azure OpenAI 打造「醫點家 AI 小幫手平台」，讓 AI 正式走入臨床文化，從技術導入進化為醫護共同創造的典範 👏最令人驚訝的是——董事長本身是放射科醫師，也親自寫 Prompt、親自做 AI 小幫手。當一位臨床領導者以身作則，代表 AI 不再是 IT 的專案，而是全院共同採用、每位醫療人員都能掌握的能力。這正是 AI 民主化（AI for Everyone in Healthcare） 的真正落地：✅ 醫護人人能用 AI✅ 醫護人人能做 AI✅ AI 成為日常臨床能力的一部分⸻🏥 落地成效亮點✅ 600+ AI 小幫手✅ 橫跨護理、醫師、藥師、行政✅ 入院護理紀錄：103.8 分鐘 → 約 6 分鐘✅ 建立 AI 小幫手「共創商城」這不只是 AI 導入，而是臨床工作模式全面升級。⸻🌐 從單院成功 → 帶動區域醫療數位轉型中山附醫提出願景：打造可複製、可共享的 AI 生態，讓中小醫院也能快速受益與 Microsoft Future Hospital DNA (Digital Native Acceleration) Program 完全契合：🔹 臨床共創🔹 AI 普及化🔹 安全的醫療雲端與資料治理⸻👏 再次恭喜中山附醫與所有親手創造 AI 的臨床夥伴，您們不只在使用 AI，您們正在引領台灣醫療的下一波創新。iThome 採訪新聞連結：- 【關鍵人物：專訪中山醫學大學董事長周英香】我們是從土壤長出來的GAI，連高層都自己動手寫提示 ： https://lnkd.in/gym7yNkW- 【從臨床痛點出發，自己的痛點靠AI自己解！】中山附醫不只要推動醫護全員AI，還要發展區域AI醫療生態系 ：https://lnkd.in/gCs-3Zxs- 【入院護理紀錄時間大省94%的關鍵】圖解醫護AI小幫手如何運作 : https://lnkd.in/gEcZdN2E- - - Chungshan Medical University Hospital × Azure OpenAI｜A New Benchmark for AI-Democratized HealthcareCongratulations to Chungshan Medical University Hospital for building the “AI Helper Platform – YiDianJia” using Azure OpenAI — bringing AI into the core of clinical culture and transforming it from a technology deployment project into a model of co-creation by healthcare professionals 👏What’s most remarkable is this:The hospital’s Chairman — a radiologist by training — personally writes prompts and builds AI assistants.When clinical leaders lead by example, AI stops being an IT initiative and becomes a capability embraced across the hospital, empowering every healthcare professional.This represents the true realization of AI democratization (AI for Everyone in Healthcare):✅ Every clinician can use AI✅ Every clinician can build AI tools✅ AI becomes a part of daily clinical practice⸻🏥 Proven Impact✅ 600+ AI assistants created✅ Adopted across nursing, physicians, pharmacy, and admin units✅ Nursing admission documentation: 103.8 mins → ~6 mins✅ Established a shared AI-tool “co-creation marketplace”This isn’t just AI adoption —it is a fundamental upgrade of clinical workflows.⸻🌐 From Single-Hospital Success → Regional Healthcare InnovationChungshan Medical University Hospital has a clear vision:Build a scalable, shareable AI ecosystem so regional and mid-sized hospitals can benefit rapidly.This aligns perfectly with the Microsoft Future Hospital DNA (Digital Native Acceleration) Program:🔹 Clinical co-innovation🔹 AI adoption at scale🔹 Secure cloud and healthcare data governance⸻👏 Congratulations again to Chungshan Medical University Hospital and every clinical innovator who built these AI assistants.You are not just using AI —you are leading the next wave of healthcare transformation in Taiwan.#AzureOpenAI #MicrosoftHealthcare #中山附醫#AI民主化 #FutureHospitalDNA #HealthcareInnovation

30

2 Comments

Pinaki Laskar

FishEyeBox AI • 33K followers

How 𝗔𝗴𝗲𝗻𝘁𝗶𝗰 𝗔𝗜 is 𝘁𝗿𝗮𝗻𝘀𝗳𝗼𝗿𝗺𝗶𝗻𝗴 𝘁𝗿𝗮𝗱𝗶𝘁𝗶𝗼𝗻𝗮𝗹 #𝘀𝗼𝗳𝘁𝘄𝗮𝗿𝗲𝗱𝗲𝘃𝗲𝗹𝗼𝗽𝗺𝗲𝗻𝘁?#𝗚𝗲𝗻𝗲𝗿𝗮𝘁𝗶𝘃𝗲𝗔𝗜 𝗮𝗻𝗱 #𝗔𝗴𝗲𝗻𝘁𝗶𝗰𝗔𝗜 𝗮𝗿𝗲 𝘁𝗿𝗮𝗻𝘀𝗳𝗼𝗿𝗺𝗶𝗻𝗴 𝗵𝗼𝘄 𝘀𝗼𝗳𝘁𝘄𝗮𝗿𝗲 𝗶𝘀 𝗯𝘂𝗶𝗹𝘁, 𝗺𝗮𝗶𝗻𝘁𝗮𝗶𝗻𝗲𝗱, 𝗮𝗻𝗱 𝗶𝗻𝘁𝗲𝗿𝗮𝗰𝘁𝗲𝗱 𝘄𝗶𝘁𝗵. 𝗛𝗲𝗿𝗲’𝘀 𝗮 𝗯𝗿𝗲𝗮𝗸𝗱𝗼𝘄𝗻 𝗼𝗳 𝘁𝗵𝗲 𝗰𝗵𝗮𝗹𝗹𝗲𝗻𝗴𝗲𝘀 𝗮𝗻𝗱 𝗱𝗶𝘀𝗿𝘂𝗽𝘁𝗶𝗼𝗻𝘀 𝘁𝗵𝗲𝘆 𝗯𝗿𝗶𝗻𝗴:🔄 Shift from Manual Coding to AI-Assisted Development:Traditional development:Requires human programmers to write, debug, and test code manually.Gen AI challenge:Tools like GitHub Copilot, ChatGPT, and Code Whisperer can now,Autocomplete code intelligently;Generate boilerplate or even complex logic;Translate between languages;Suggest fixes or optimizations;📌 Impact:Reduces the need for repetitive coding;Lowers the barrier to entry for non-developers;Raises expectations for faster development cycles;🤖 Autonomous Software Agents:Traditional development:Applications run predefined logic and require user commands or inputs.Agentic AI challenge:AI agents can,Plan, reason, and make decisions independently;Interact with APIs, databases, and other software without being explicitly programmed for each task;Adapt and learn from the environment;📌 Impact:Moves from static systems to dynamic, adaptive ones;Introduces a new development paradigm: "specify the goal, not the steps";Challenges existing architectures which are rigid and procedural;🛠️ New Development Paradigms:Traditional software:Built using structured design, with heavy focus on deterministic logic and clear data flow.AI Agents approach:More probabilistic and data-driven;Can evolve through fine-tuning and prompt engineering;Uses "natural language" as a development interface (e.g., prompt-based programming);📌 Impact:Engineers must now also be prompt designers or data curators;Testing and debugging becomes less about syntax errors and more about behavior alignment;Version control of prompts, training data, and models becomes crucial;📦 Challenges to DevOps & Maintenance:Traditional model:CI/CD pipelines manage code changes, testing, and deployments.Agentic AI complications:Code is partly generated dynamically — versioning is less straightforward;Models might change behavior after updates or retraining;Monitoring and auditing AI behavior requires new tools (e.g., for explainability, fairness, or hallucination detection);📌 Impact:Observability and debugging tools need to evolve;Governance, compliance, and reproducibility become more complex;

269

18 Comments

Coupang

247K followers

Coupang is reimagining the shopping experience with the goal of wowing each customer from the instant they open the Coupang app to the moment an order is delivered to their door. The role of machine learning is becoming increasingly important in this innovation. Coupang's machine learning technology optimizes product search, pricing, inventory management, and logistics processing, and improves our customers' lives exponentially better. For more details, visit Coupang's Engineering Blog. ✅ Engineering Blog: https://lnkd.in/eRydERW5 #Coupang #CoupangCareers #CoupangTech #Tech #Engineer #DataEngineering

90

Effie GUO

VCBeat Health • 7K followers

China's AI healthcare agent market is exploding.Here's the complete ecosystem map:1. UPSTREAM: The Foundation LayerInfrastructure Computing:• NVIDIA,AMD,Intel Corporation,, Iluvatar CoreX 天数智芯 powering the compute• Google Cloud, Amazon Web Services (AWS) Cloud, Microsoft Azure, Tencent Cloud, Huawei Cloud, Alibaba Cloud scaling deploymentAlgorithm Frameworks:• Large models: DeepSeek AI, Google, OpenAI, Zhipu AI, AQ Health Service• Development Frameworks: Microsoft Azure, NVIDIA, Weaver BPM Alibaba Cloud, OpenAI2.  MIDSTREAM: The Application LayerPre-diagnostic Phase (12 companies):IFLYTEK Healthcare, Baidu Healthcare, Winning Health Technology Group Co.,Ltd , Ngarihealth Technology Co.,Ltd ,北京左医科技有限公司, Future Doctor, TencentHealth, PingAn Medical & Healthcare, JD Healthcare & JD Lifescience, 美年大健康产业(集团)有限公司 AQ Health Service China Resource Sanjiu Medical & Pharmaceutical Co., LtdHospital Management (11 companies):SYNYI AI, 智云健康, Silicon Intelligence, Inc. Huimei Healthcare XRAYBOT, Ning Tang Healthcare, Wanren Wisdom Intelligent Technology (Beijing) Co., Ltd. , Inspur Group, Tairex, 清华大学智能产业研究院万达信息股份有限公司,Huawei, Ruijin Hospital Affiliated to Shanghai Jiao Tong University, School of Medicine Fangzhou Inc. (方舟健客 6086.HK) Medical Imaging Diagnosis (9 companies):Neusoft Medical, Airdoc Technology Inc , Medical AI, Infervision，HealthGPT+, Linkingmed, Huiying Medical Technology Co.,Ltd. ,深睿医疗Deepwise, Yidu Tech Inc. Clinical Decision Support (11 companies):iFlytek Healthcare, Winning Health Technology Group Co.,, 北京左医科技有限公司, 金域检验, Beijing Digihit Technology Co.,Ltd., 睿心医疗,Future Doctor, HUAWEI, West China Hospital, Shanghai Runda Medical Technology Co.,Ltd., Zhipu AI3.  DOWNSTREAM: The End UsersTo B - Medical Institutions:• ZhongShan Hospital, Fudan University• Fudan University Shanghai Cancer Center • West China Hospital• Ruijin Hospital Affiliated to Shanghai Jiao Tong University, School of Medicine• Shenzhen University Affiliated South China Hospital To C - Patients:• Online Consultation: PingAn Medical & Healthcare,, Deep Spring Healthcare好大夫在线-互动峰科技(北京)有限公司, Spring Rain Software • Health Management: Health Hope, @Keep, IVYLINK The Pattern: Infrastructure first → Algorithms → Applications → End users. This is coordinated ecosystem building, not fragmented innovation. 200+ companies vertically integrated from chip to patientP.S. What else I miss? Eng. Muhammad Mudassar (محمد مدثر) , SMIEEE, MBA, CPHIMS, PMP| Jon Warner| Bron Kisler | Brian de Francesca | Rémy LEVASTRE#ChinaHealthTech #AIHealthcare #HealthTechEcosystem #MedicalAI #MarketMap—Enjoy this? ♻️ Repost it to your network and follow Effie GUO for more.

186

28 Comments

Gayathri G

OptiSol Business Solutions • 4K followers

Alibaba Group’s #Qwen team just dropped FP8 builds of Qwen3-Next-80B-A3B in both Instruct and Thinking flavors.Key details:⚡ FP8 (block-128) → cuts memory + bandwidth without giving up scale🧩 Hybrid-MoE → 80B total, ~3B active (512 experts: 10 routed + 1 shared)📏 Context → 262K native, ~1M validated via YaRN🧠 Thinking build → defaults to <think> traces, pairs with a reasoning parser🔮 Features → multi-token prediction + serving commands for latest sglang/vLLM nightlies📊 Benchmarks shown are from BF16 versions, so FP8 performance needs re-validation.✅ Licensed under Apache-2.0 → open for commercial + research use.Another step in pushing massive reasoning-capable MoEs onto commodity GPUs.Link : https://lnkd.in/gTtJw9Mh

17

1 Comment

Liu Yu-Wei

Accelerate Private Machine… • 2K followers

[ DevOps in AI Agent ] Google ADK AI Agent Evaluation 概念介紹 (1/2)> The English version is below.> 此系列因文章較長，我將會將文章拆成「功能介紹篇」與「技術實作篇」，再請大家閱讀並且按鼓勵給予支持。當今天將 AI Agent 開發完成後，依照軟體開發的流程，單元測試（Unit Test）與整合測試（Integration Test）扮演著確保系統穩定與可靠的關鍵角色。然而，大型語言模型 AI Agent 的出現，帶來了全新的挑戰。由於模型本身具備機率性，相同輸入在不同情境下，從使用第三方工具到回覆結果，皆有可能產生略有差異的輸出。因此，對這類系統而言，「通過／失敗」這類二元判斷並不足以反映 AI Agent 的整體表現。本文將深入介紹 Google ADK AI Agent 開源工具中，是如何設計 Evaluation 這項功能，以及說明它如何協助開發者在機率性與可變性的環境中，依然維持對 AI Agent 的品質掌控與開發信心。Link: https://lnkd.in/gKk6QZFg===[ DevOps in AI Agent ] Introduction to Google ADK AI Agent Evaluation Concept (1/2)> Due to the length of this series, I’ve divided the article into two parts — “Feature Overview” and “Technical Implementation.” Please take some time to read them and show your support with a clap.Once an AI Agent has been developed, following standard software development practices, **unit testing** and **integration testing** play a critical role in ensuring the system’s stability and reliability.However, the emergence of **large language model (LLM)–based AI Agents** introduces new challenges. Because these models are inherently probabilistic, the same input under different conditions — from invoking third-party tools to generating responses — may produce slightly different outputs. As a result, binary judgments like “pass” or “fail” are insufficient to capture the full performance picture of an AI Agent.This article explores how the **Evaluation** feature is designed within **Google ADK’s open-source AI Agent toolkit**, and explains how it helps developers maintain quality control and confidence in their AI Agent’s performance, even within an environment of variability and probabilistic behavior.Link: https://lnkd.in/gKk6QZFg

9