Insights, tutorials, and news from our team
Recently, Jimmy Song (Song Jingchao), a senior technical expert in cloud-native and open-source fields, officially joined Dynamia AI as VP of Open Source Ecosystem, helping Dynamia AI in AI native infrastructure ecosystem construction and global developer collaboration layout.
Shanghai Dynamia Intelligence Technology Co., Ltd. ('Dynamia AI'), an AI startup focused on heterogeneous compute scheduling and virtualization, recently completed tens of millions of RMB in angel financing, led by Fosun RCC with strong participation from Zhuopu Investment, seed investors, and industrial partners.
On December 27, 'Efficiency Over Compute Power: HAMi Meetup Beijing Station' concluded successfully in Beijing's Haidian District. Nearly 100 technical partners gathered to discuss heterogeneous compute virtualization, scheduling strategy evolution, and AI business efficiency improvements.
Xinference natively supports HAMi vGPU through Helm Chart, enabling multi-model inference platforms to achieve secure GPU sharing, fine-grained quotas, and unified governance. This article details how to enable HAMi vGPU with one click and implement best practices for cost reduction in production environments.
HAMi v2.7.0 introduces an extended ResourceQuota mechanism to address limitations of native Kubernetes ResourceQuota in GPU and heterogeneous computing scenarios. This article provides an in-depth code analysis of how HAMi solves the two major pain points of resource correlation and dynamic resources, offering precise GPU resource quota management for multi-tenant environments.
HAMi v2.7.0 introduces topology-aware scheduling for NVIDIA GPUs. This article provides an in-depth code analysis of HAMi's design and implementation principles for topology-aware scheduling, explaining how intelligent scheduling precisely deploys compute tasks on the most tightly connected GPU combinations.
HAMi v2.7.0 is here! Featuring full support for Kunlun XPU, Enflame GCU, AWS Neuron, and MetaX, enhanced scheduler optimizations, expanded application ecosystem integration, improved WebUI functionality, and thriving community growth.
Follow this battle-tested checklist to prepare IAM, install cluster add-ons, and roll out the Dynamia AI Platform on Amazon EKS in about an hour.
Explore a practical 2×2 framework for GPU scheduling in Kubernetes: combining node and GPU-level binpack/spread strategies. Learn how HAMi enables device-aware scheduling beyond native K8s capabilities, with hands-on examples of four distinct patterns that balance cost efficiency, GPU availability, and performance for AI workloads.
Get the latest technical articles, tutorials, and HAMi community updates.