Platform

One control plane for edge AI capacity, power intelligence, and secure inference APIs.

UmamiEdge is designed to begin as software, validate with controlled GPU pilots, then scale into a global regional compute network.

Site Hosts

Telecom towers · universities · banks · hospitals · solar mini-grids

Power Intelligence

Grid + solar + battery telemetry, demand windows, thermal envelope

Compute Nodes

GPU/accelerator servers, secure boot, local model cache

UmamiEdge Control Plane

Node registry, workload router, tenant policies, billing meter

AI Customers

APIs for local-language LLMs, document AI, vision, SOC copilots

Modules

MVP components.

Onboard GPU nodes, site hosts, hardware profiles, power limits, network status, and maintenance windows.

Route workloads by latency, capacity, model availability, data-residency policy, and energy state.

Shift lower-priority jobs into solar/battery windows and throttle workloads during constrained power events.

API keys, tenant isolation, signed node heartbeats, audit logs, model access control, and incident flags.

Track tokens, GPU seconds, node revenue share, site-owner credits, and enterprise invoices.

Expose curated multilingual models and APIs for document AI, cybersecurity, agriculture, tourism, industry, and customer operations.

Roadmap

Phase 1

0–60 days

Phase 2

60–120 days

Phase 3

120–240 days

Phase 4

9–18 months