Senior Solution Architect at Lenovo, focused on AI-driven operations, AIOps, multi-cloud management, and sustainable computing.
I am a solution architect in enterprise IT operations and management. My work focuses on applying AI and machine learning to intelligent operations, digital transformation, AI-assisted monitoring, and hybrid multi-cloud management.
Everything I touch, from me outward: work, projects, talks, patents, awards, content and the technologies that connect them.
Hover or tap a node to see what it is and how it connects; click (tap twice on mobile) to open it. Drag nodes, use + / − to zoom, drag the background to pan.
Only work from the Lenovo era (December 2024 onward): AI products, enterprise AI platforms, agent skills and the infrastructure behind them. Eight projects, Oct 12 – Nov 12, 2026, each with dates and the tech used.
Own · Notion project page
A bilingual (Chinese / English) hub for the day's top AI news: each story comes with key points, primary sources and a heat score, and the archive can be filtered by company, technology, impact and date. Stories are picked from what is most discussed on X and checked against primary sources by an automated agent pipeline; optional narrated explainers accompany each edition. Article picks brings together enterprise AI articles with bilingual summaries, original thumbnails, and links to the sources.
Screenshot: aibriefhub.com Article picks
Systems that run on their own every day.
Moved the videos to object storage behind a dedicated HTTPS CDN domain, aibriefhub.com, with a fixed naming scheme, upload verification and a machine-readable “latest” pointer, so an enterprise AI assistant skill can fetch the newest episode on its own.
Own guide graphic
An automated executive AI intelligence system that monitors public enterprise-AI signals, scores them for strategic impact, and pushes structured briefs into a chat channel. A semi-automated version is already running day to day; the design covers real-time alerts, a daily intelligence brief and trend analysis.
Own · Notion project page
Ran the enterprise AI assistant through a high-stakes leadership meeting: discussion from eighteen tables and the main stage was turned into customized summaries and syntheses in six formats, with zero support incidents and a round-the-clock war room behind it.
AI brought into the tools large teams already use.
Shipped a run of releases for an internal AI workspace: a Chinese/English management console with role-based account management and a verified offline upgrade path (Oct 23), page-visit analytics with sign-in attribution (Oct 20), and a standalone meeting dashboard (Oct 19). Built with AI coding agents.
Own · Notion project page
Brought an internal enterprise AI assistant into Teams group chats as a custom tab with single sign-on working end to end, and piloted a self-hosted agent as a Teams bot: app manifest packaging, Azure Bot registration and a Cloudflare Tunnel webhook. The bot install is awaiting tenant approval.
Logo: Microsoft, via Wikimedia Commons
Designed machine-learning models that predict server performance and failures and raise alerts automatically, and wired ServiceNow and PagerDuty into Slack-based ChatOps so incidents are routed by severity to the right runbook or engineer. Used Maximo Visual Inspection on OpenShift for image-based asset and data center monitoring.
Slide: own SREcon22 APAC deck (USENIX)
Packaged skills that let AI assistants do real work.
Rebuilt the knowledge pack behind a concierge assistant for a large technology event: a 202-page master deck processed into a searchable PDF, registration lookup verified, and the build reviewed by a second coding agent before release.
An assistant skill that turns uploaded templates and guest lists into ready-to-send Outlook invitation drafts with per-guest CC, plus a short summary report. Follows v1.1.2 (Sep 30), which fixed recipient resolution.
Frame: own training video
The machines and networks behind the work.
Set up an NVIDIA DGX Spark as a remote coding workstation that AI agents can drive: Codex CLI reachable over a secured jump host and proxy, used to build skills and to have one agent review another's work.
Photo: NVIDIA
Runs a self-hosted, Tailscale-compatible control plane that joins a laptop, the DGX Spark, a cloud VPS and a home-lab VM into one private mesh. Cleaned up idle clients and stale firewall rules on Oct 23, and added public HTTPS ingress with no open ports through Cloudflare Tunnel on Oct 25.
Logo: headscale project (BSD-3-Clause)
Data-driven efficiency for real production data centers.
Technical and project lead for an end-to-end platform in IBM's Beijing systems center: infrastructure observability, application performance management, active resource management and dynamic workload scheduling, measured with about 100 sustainability metrics across four dimensions. A volunteer innovation project with more than 80 colleagues from 8 countries.
Diagram: own (Notion)
Built and operated a high-availability data center at an annual PUE of 1.25 while keeping service SLAs and SLOs, through airflow redesign, infrastructure monitoring and data-driven efficiency tuning. Awarded Best Data Center Efficiency in China (Datacenter Dynamics GCG, 2012).
Slide: own SREcon22 APAC deck (USENIX)
Delivered the sustainability demo and live demo with the software team for the global launch and promotion events of IBM LinuxONE Emperor 4. The demo was published on the IBM LinuxONE website and the product won a 2022 SEAL Sustainable Product Award.
IBM LinuxONE sustainability · SEAL Awards 2022
Image: IBM LinuxONE website
IoT sensors monitor UPS batteries in real time; mechanism models and AI models analyze battery health, detect anomalies and recommend proactive maintenance. Monitoring and analysis run in production for the systems center's SREs; end-of-life prediction and waste management were proven in PoC.
Photo: SREcon23 APAC slides (USENIX)
Conference sessions and published research.
Participated in the conference, with a proposed session on full-stack sustainability optimization of cloud-native workloads with open-source software in real-world production: challenges of heterogeneous infrastructure, real-time data analysis, AI-driven insights and automation.
Own speaker profile (Sessionize)
Co-presented with Chen Wang and Fan Jing Meng at the Linux Foundation's Open Source Summit.
Thumbnail: The Linux Foundation (YouTube)
Co-presented with Marcelo Amaral and Fan Meng: a measurement system for sustainable computing and a live demo in a heterogeneous, hybrid cloud data center.
Thumbnail: The Linux Foundation (YouTube)
Presented by IBM colleagues David Cesarano and Fanjing Meng: how IoT sensing and AI models extend battery life and cut waste in a production data center.
Slide: USENIX SREcon23 APAC
Concepts and technical challenges of a full-stack sustainability optimization platform, its measurement standards and testing approach, with an end-to-end live demo.
Slide: own SREcon22 APAC deck (USENIX)
Co-author. Wrote the chapter on the design and results of sustainable computing at IBM China Systems Center, in the blue paper jointly published with the China Academy of Information and Communications Technology.
Photo: CAICT (WeChat)
Live demos and panel discussion on sustainability strategy, green IT and sustainable IT infrastructure.
Image: IBM Academy of Technology (Notion)
US patent application US 2024/0241885 A1 (app. 18/155,242). Scheduling data collection from the temporal pattern of data generation, resource pricing and collection goals. Inventors include Hua Ye; assignee IBM. LinkedIn lists this under Patents; Google Patents shows a published application, not a grant.
Drawing: Google Patents
US patent application US 2024/0241870 A1 (app. 18/154,985). Builds resource-aware dynamic operational data collection and analysis plans for technical environments. Inventors include Hua Ye; assignee IBM. LinkedIn lists this under Patents; Google Patents shows a published application, not a grant.
Drawing: Google Patents
US patent application US 2024/0403181 A1 (app. 18/204,707). Detects and predicts storage-device data anomalies. Inventors include Hua Ye; assignee IBM. Matches Notion disclosure P202204601.
Drawing: Google Patents
US patent application US 2025/0117551 A1 (app. 18/482,094). Estimates power used when addressing application requests with machine-learning models. Inventors include Hua Ye; assignee IBM. Matches Notion disclosure P202204600 / application energy monitoring.
Drawing: Google Patents
Listed on LinkedIn Patents and on the Notion Invention Patent page as IBM disclosure CN8-2012-0791 (also written CN820120791). No public CNIPA or USPTO publication was found under that title with inventorship confirmed, so this is shown as a disclosure, not a granted patent.
Source: Notion Invention Patent page · LinkedIn
IBM LinuxONE, part of the IBM Full-Stack Sustainable Computing Solution, recognized as one of 13 Outstanding Cases of 2023.
Image: own (Notion)
Recognition for the full-stack sustainable computing project's multi-technology architecture design and validation.
Photo: own (Notion)
First prize, Best Project Award in the GCG Innovation Factory, with the Excellent Project Leader Award; winner of the Resilient IT Infrastructure theme in the Sustainability Challenge 2022.
Slide: own (Notion)
IBM Technical Specialist Profession, Expert (Level 2); Enterprise Design Thinking; Data Science Foundation; Developing an AI/ML Data Strategy (2025); Lenovo – Managing in the AI Era (2025).
Badge: own certification
Short explainer videos I make and publish as 星尘, including “DiagramGPT 告别手动画图” (DiagramGPT: no more drawing diagrams by hand), “神奇的动作捕捉” (motion capture), “与动物对话” (talking to animals), “极端天气的成因与预测” (causes and prediction of extreme weather) and “网球 数据大揭秘” (tennis by the numbers). 1,533 likes and saves.
Own · Xiaohongshu
Short videos published on WeChat Channels (微信视频号) as 星尘stardust.mike. Search the account name in WeChat to watch.
Own · video cover
Travel and city vlogs, including Angkor Wat, a New Zealand road trip, a US East Coast road trip, Kyoto and the first snow in Beijing.
Own · YouTube
Photography is a long-running passion outside work.
Photo: Ye Hua