@HAMiProject
Account based inSingapore!
About this account
- Account based in
- Singapore
- Connected via
- China App Store
X says this location may be affected by a proxy or VPN.
Account-level information from X, not a live location or the device used for a specific post.
HAMi is a @CloudNativeFdn open-source GPU virtualization middleware for sharing, isolating, and scheduling heterogeneous accelerators on Kubernetes.
Earth
Joined July 2026
- Tweets49
- Following51
- Followers31
- Likes42
Thanks to the @AI_TCG_CNCF for having us today!
Mengxuan Li, Dynamia Co-founder & CTO and HAMi author, showed how HAMi gets more out of heterogeneous AI clusters: GPU sharing, isolation & topology-aware scheduling on K8s.
Thanks to everyone who joined.
Great explanation @HowDevelop
So impressed with the Opus 5.5 explainer videos - Here's another one - Serve Models from a @Kit_Ops ModelKit on @HAMiProject
Detailed explanation of how Your model is packaged as a versioned ModelKit. KitOps pulls it from an OCI registry and unpacks it inside the Pod. HAMi schedules a controlled GPU share. SGLang loads the model locally and serves an OpenAI-compatible API. One clear path from model artifact to working inference. KitOps and HAMi.
Happy Mid-Autumn Festival from the HAMi community!
Wishing our community around the world a joyful season of connection and togetherness.
#HAMi #OpenSource #MidAutumnFestival
Want to see how HAMi manages heterogeneous GPU resources on Kubernetes?
Join the @CloudNativeFdn AI Technical Community Group session on September 29 to see how HAMi approaches topology-aware scheduling and GPU sharing across different vendors.
ocgroups.dev/cncf/group/6auh…
How does HAMi handle heterogeneous GPUs on Kubernetes?
Join this CNCF AI community session on Sep 29 for a practical look at topology-aware scheduling, GPU sharing, and heterogeneous AI infrastructure.
ocgroups.dev/cncf/group/6auh…
#HAMi #Kubernetes #CNCF
Project HAMi retweeted
Boosting dev capacity 15x on the same hardware sounds impossible until you rethink GPU scheduling.
CETC Cloud used Kubernetes + HAMi to scale from 2 to 30 concurrent dev Pods on an 8-GPU device with production AI.
cncf.io/case-studies/cetc-cl…
Ready to optimize your AI infrastructure? Help build and refine GPU virtualization with Kubernetes and HAMi on GitHub: github.com/Project-HAMi/HAMi
#Kubernetes #AI
Your GPU really costs more than a car.
Also your GPU: 12% busy.
HAMi lets multiple Kubernetes workloads share one GPU, with per-pod limits on GPU memory and GPU compute. Same hardware, way less idle silicon.
60-second explainer, What's your cluster's real utilization?
Welcome Aditya Raut to the HAMi community!
github.com/adity1raut
With 15 merged PRs across scheduler, device plugin, multi-vendor backends, and testing, Aditya has already made a strong contribution to HAMi.
Thanks for your work, and welcome aboard!
#HAMi #OpenSource #GPU
New @CloudNativeFdn case study: Sangfor uses HAMi with Kubernetes & Volcano to achieve 3x+ GPU utilization and 8x+ model density per GPU.
cncf.io/case-studies/sangfor…
#HAMi #CNCF
Welcome @YiXinGroup as a contributor to HAMi!
Their Training and Inference Platform team has contributed production-proven improvements in GPU resource defragmentation, full-GPU workload optimization, and monitoring.
Thanks for contributing back to the HAMi community.
🤖 Made with AI
HAMi in Hyderabad (India)!
Our Ambassador @maishivamhoo introduced @HAMiProject to the engineers at Eurth Techtronics Pvt. Ltd , showing how GPU sharing on Kubernetes powers efficient AI infrastructure.
Thanks for hosting us!
#HAMi #Kubernets #GPU #AIInfrastructure
Project HAMi retweeted
when you Install HAMi, you get HAMi-core, a library (libvgpu.so) that sits inside your pod. HAMi tells it your limit through an env var (CUDA_DEVICE_MEMORY_LIMIT). When your job asks the GPU for memory, the library checks under the limit it passes, otherwise fails.
New HAMi lab run SGLang on shared GPUs
Deploy SGLang on Kubernetes with HAMi: fractional GPU allocation, memory/compute limits, Qwen3-1.7B behind an OpenAI-compatible API, and verified GPU isolation.
Thanks to @rudrakshkarpe & @HowDevelop for contributing!
project-hami.io/tutorials/la…
New HAMi Lab: Run SGLang on HAMi GPU Shares
Deploy SGLang on K8s with fractional GPU memory & compute limits, and verify HAMi’s GPU isolation in practice.
Thanks @rudrakshkarpe and @HowDevelop for contributing!
project-hami.io/tutorials/la…
#HAMi #SGLang #Kubernetes #AIInfra
What a week at @KubeCon_ China 2026!
Our first KubeCon China as a @CloudNativeFdn Incubating project: 2 keynotes, a lightning talk, a tech session & a full-day booth in Shanghai.
Thanks to everyone who talked GPU infra with us. Full recap
project-hami.io/blog/kubecon…
What a day at the @HAMiProject booth at @KubeCon_ China! 🎉 Non-stop visitors and great conversations from start to finish.
Huge thanks to Mengxuan Li, @jimmysongio, @fishmanticoder, Chris Gao, Xiao Zhang and Jifei Wang, for running the booth 🙌
#KubeCon #CloudNative #HAMi
China Merchants Bank manages nearly 10,000 accelerators for AI training & inference using Kubernetes, Kueue, KEDA, HAMi, Fluid & Prometheus.
HAMi enables fine-grained GPU sharing and scheduling, improving utilization and reducing costs.
cncf.io/case-studies/china-m…
Meet our first HAMi Ambassadors!
Last night at HAMi Community Night, @jimmysongio announced our inaugural cohort of HAMi Ambassadors for the 2026–2028 term:
Mesut Oezdil, 王天青, Xueduan Li, @maishivamhoo , @SaiyamPathak
Congratulations to all five!
project-hami.io/community
HAMi is coming to KubeCon China 2026!
Dynamic MIG, GPU virtualization at scale, and heterogeneous AI infrastructure, see what the HAMi community will be sharing in Shanghai.
project-hami.io/blog/hami-at…
#HAMi #KubeCon #AIInfra
