Run AI applications with the Enterprise Edge AI platform Palette Edge

Why edge is critical for AI workloads

AI applications will increasingly be deployed in edge locations in order to get closer to the data, the process and the user. This is essential for:

Performance

AI inferencing workloads usually runs in real-time and therefore can’t tolerate latency for data to be transmitted to the cloud for analysis and action.

Availability

Mission-critical AI workloads such as in connected cars must remain active even if the edge location loses connection to central clouds or data centers.

Cost

Video, audio and other data sources may be data intensive therefore too costly to move over the network, with cloud ingress, egress and storage costs on top.

Security

Data gathered for AI inference may be sensitive, meaning strong compliance limitations on moving the data over the network and into the cloud.

Key capabilities for edge AI

Spectro Cloud’s Palette Edge solution now features Edge AI capabilities, designed to help enterprises like yours embrace the potential of AI at the edge. You’ll be able to:

Deploy and manage your edge AI stacks to edge locations at scale

To run AI workloads, each edge location needs servers with an OS, Kubernetes, integrations such as security and observability tools, a model serving framework such as Kubeflow, and one or more AI models deployed and configured. Doing this manually, hundreds or thousands of times, is cost prohibitive — especially if you need an expert to visit each site in person.

Palette’s Edge AI makes it easy, with:

Repeatable ‘blueprints’ of the full AI and infrastructure stack, from OS to Kubeflow and Seldon, using Palette’s Cluster Profiles. You can enforce your ‘desired state’
Low- and no-touch provisioning of all kinds of edge devices, including Arm and x86 architectures, with direct bootstrapping of bare metal devices
Unique resilience and scalability with its patented decentralized architecture

Update your edge artificial intelligence models frequently without risk of downtime

Competitive innovation comes from non-stop incremental improvements in your AI models. How can you push new code daily, and roll back safely with full control? Palette’s EdgeAI enables you to:

Access AI model control via your own private repositories and from Hugging Face
Support safe experimentation with declarative, version-controlled, GitOps-able model configurations and A/B testing
Make controlled changes fast, with zero downtime rolling upgrades that you can apply to single locations, groups of locations, or even your entire state of thousands of edge devices if you need to — executed in parallel
Control risk with canary model deployments and easy rollback to previous model versions
Monitor performance across multiple edge locations with native observability dashboards

Secure critical intellectual property and sensitive data with ai on the edge

Any edge computing has risks — and with edge AI, the stakes are high. The models themselves may be your critical IP, and they may be deployed in vulnerable edge locations that you don’t own. The data they gather and analyze may be sensitive and covered by compliance requirements.

Palette’s Edge AI solves for security with:

Strict standards compliance. Palette is available with FIPS cryptography across all elements of its architecture. Of course it’s also compliant with ISO 27001, SOC 2 and other standards, and has passed infosec audits with some of the most stringent organizations
Multi-layered security controls, from immutable OS images and persistent data encryption to verified secure boot, cluster hardening, SBOM and other security scans, and zero trust with granular RBAC. You can trust that every node that you bring online is trusted and secure.
Full deployment flexibility. You can build your clusters with your preferred OS, security integrations and versions. You can deploy clusters air-gapped and even deploy your own instance of Palette in an air-gapped environment.