Now in private alpha — autonomous edge in 12 regions

Autonomous Edge
Infrastructure for the AI Era

AI-native CDN, storage, email, and observability — orchestrated in real time by token-less autonomous agents across a self-healing global network.

SFOORDJFKLHRFRADXBSINSYDGRUMEXCAIBOM
<10ms
Edge Latency
100ms
Global Config Sync
75%
Lower Ops Overhead
0
AI Token Fees
The Problem

The cloud was never built for AI-scale operations.

Modern infrastructure forces teams to choose between speed, cost, and autonomy. We refused.

Fragmented Stack

CDN, storage, email, and observability scattered across vendors with brittle glue code.

Hyperscaler Pricing

Egress fees, AI tokens, and per-request markups that scale faster than your revenue.

Token-Based AI Costs

Every inference billed per token — unpredictable costs that punish high-throughput workloads.

Manual Operations

On-call engineers tuning configs, draining nodes, and chasing alerts that should self-resolve.

Slow Config Propagation

Minutes to roll out a rule change globally — an eternity at modern request volumes.

Reactive Observability

Dashboards that explain failures after they've already cost you customers.

The Platform

One autonomous platform.

Six primitives, one control plane. Built to operate themselves at planetary scale.

AI-Native CDN

Adaptive routing trained on your traffic shape.

Object Storage

Hot, durable, and globally replicated by default.

SMTP Edge

Deliverable email with edge reputation in real time.

AI Agents

Token-less, embedded — they run the platform.

Realtime Config

Propagate any rule globally in under 100ms.

Observability

Predictive metrics, traces, and logs unified.

Performance

Built for extreme scale.

C++ Seastar core. Shared-nothing per-CPU architecture. RAM + NVMe optimized. Designed to saturate modern hardware, not waste it.

1M+
req/sec
Per Edge Node
<10
ms
P50 Edge Latency
<100
ms
Global Config Sync
75%
Ops Overhead Reduction
nomad.live / control-plane
live
Edge Throughput
847,392 req/sec
P50
6ms
P99
14ms
Errors
0.001%
AI Agents

Token-less AI infrastructure.

Our agents are embedded in the data plane — not bolted on as chatbots. They make decisions in nanoseconds, not API calls.

Autonomous Routing
Real-time path selection based on regional load, reputation, and cost.
Self-Healing
Drain, reroute, and replace failing nodes before SLOs are touched.
Cost Optimization
Continuous tier and region rebalancing tied to your access patterns.
Predictive Scaling
Capacity arrives before demand. No cold starts, no thundering herd.
Anomaly Detection
Statistical & embedding-based outlier detection across every plane.
Automated Rollback
Bad deploys reverted globally in under 100ms with full audit trails.
The Autonomous AI Engine

A single fabric. Six intelligent planes.

NomadCDN is engineered as one self-operating system — an orchestrated control plane, a distributed data plane, an agentic AI ecosystem, and continuous self-healing, cost, and learning loops that make the platform smarter every request.

01 — Control Plane

The orchestration brain.

Apache Airflow runs autonomous DAGs that continuously plan, schedule, and propagate global policy. Every decision is auditable, every change reversible in under 100ms.

Autonomous AI DAG Workflows
Apache Airflow
Routing Intelligence
DAG · live
Cost Optimization
DAG · live
Infra Intelligence
DAG · live
Hardware Optimization
DAG · live
GeoDNS Management
DAG · live
Security Intelligence
DAG · live
Enterprise Workflow
DAG · live
Airflow Core Services
Scheduler
DAG Parser
Executor (K8s)
Metadata DB
XCom / Object Store
RBAC & Permissions
Web UI & API
AI Optimization Engines
always-on
Routing AI
Traffic prediction, path optimization
Cost AI
Transit cost, ISP optimization
Infra AI
Health scoring, capacity forecasting
Hardware AI
Resource placement & rebalance
Security AI
Threat detection, mitigation
GeoDNS AI
Latency prediction, geo-routing
Policy & Decision
Policy synthesis, what-if analysis
02 — Distributed Data Plane

Edge runtime, planet-scale.

Each NomadCDN PoP runs a Seastar C++ shared-nothing runtime. RAM-hot, NVMe-warm, GPU-ready — saturating modern hardware instead of wasting it.

Edge Ingress & Traffic
Anycast IP (v4/v6)
QUIC / HTTP/3
L4–L7 Load Balancer
Rate Limiting & Shielding
WAF / DDoS
Seastar C++ Core Services
Connection Manager
Request Router
Cache Engine
Object Manager
Stream Manager
TLS 1.3 / QUIC
Compression
AI Inference Runtime
eBPF Telemetry
Local Data & Compute
RAM Cache (hot)
NVMe Cache (warm)
RocksDB (metadata)
Local Vector DB (semantic)
GPU / NPU (inference)
Distributed Metadata
ScyllaDB

Sub-millisecond key lookups across regions. Backs routing tables, cache indices, ACLs, and agent state with eventual quorum consistency.

us-east
eu-west
ap-south
sa-east
af-cape
oce-syd
Object Storage
erasure-coded

12+4 Reed–Solomon coding. Multi-region replication, cross-PoP sync, and a consistency engine driven by the same AI policy plane.

Multi-region replicationactive-active
Cross-POP sync<100ms
Consistency enginetunable
03 — Autonomous Edge AI Agentic Ecosystem

Token-less agents at every hop.

Agents speak MCP (Model Context Protocol) for tool use and A2A (Agent-to-Agent) for peer coordination. They run inside the data plane runtime — no external API calls, no per-token bills.

Protocols & Interfaces
open
MCP
Model Context Protocol — tool & context invocation.
A2A
Agent-to-Agent — peer coordination & negotiation.
Enterprise Surface
Copilots & Ops Agents
FinOps & Security Agents
Support & Workflow Agents
CRM / ERP / ITSM / CI/CD
Edge Control Agents
per-POP
Health Monitoring
Continuous SLO probing of every subsystem.
Local Optimization
Per-PoP cache, route, and inference tuning.
Policy Enforcement
Applies global policy with local context.
Autonomous Actions
Drain, scale, reroute — without paging humans.
A2A Interface
Negotiates load and failover with peer PoPs.
AI Inference Runtime
Embedded models — no external token API.
04 — Self-Healing

Failures resolve before pagers ring.

A closed loop between telemetry, AI engines, and edge agents drains, reroutes, and rebuilds — all under 100ms of customer-visible impact.

01
Detect
eBPF telemetry + anomaly models flag degradation in <1s.
02
Diagnose
Routing AI correlates symptoms across PoPs and planes.
03
Drain & Reroute
Traffic shifts via Anycast; sessions migrated transparently.
04
Recover
Failed nodes rebuilt; replication restored to N+2.
05
Verify
Synthetic probes confirm SLOs before normal routing resumes.
<1sAnomaly detection
<100msFailover propagation
N+2Replication invariant
05 — Cost Efficiency

Every byte routed at the cheapest viable path.

The Cost AI engine continuously rebalances transit, peering, hardware tier, and storage class against your real access patterns — without trading SLOs.

Optimization Loops
continuous
Transit & Peering Arbitrage
Per-prefix BGP economics evaluated every minute.
Tier Rebalancing
Objects migrated between RAM ↔ NVMe ↔ erasure-coded cold storage by access heat.
Hardware Placement
Workloads bin-packed across CPU / GPU / NPU per cost-per-token-equivalent.
Egress Elimination
Zero per-GB egress fees — included by design.
Customer Outcome
vs hyperscalers
30–70%
Lower total infra cost
$0
AI token fees
$0
Egress surcharges
Flat
Predictable pricing
Hyperscaler billNomadCDN
avg. customer · 62% reduction
06 — Self-Learning

A platform that gets smarter every request.

Telemetry flows back through a feature store into model training. New policies are A/B-tested on shadow traffic, then propagated globally — closing the loop in minutes, not quarters.

Data & Telemetry Pipeline
continuous learning
Edge Telemetry
Metrics · Logs · Traces
Event Bus
Kafka / NATS
Stream Processing
Flink / Kafka Streams
Time-Series DB
Prometheus / Victoria
Data Lake
S3 / Iceberg
Feature Store
Vector DB
AI/ML Training
Models → Policy
Observe
Every request becomes a training signal — anonymized at the edge.
Synthesize
Policy & Decision engine proposes candidate strategies.
Roll Out
Shadow → canary → global, with instant rollback if SLOs drift.
Autonomous Content Fabric · Live Console

One pane of glass for the entire fabric

The NomadCDN Control Plane unifies edge, POP, and region telemetry into a single autonomous operations console — with AI making routing, caching, and capacity decisions in real time.

NomadCDN
NomadCDN | Autonomous Content Fabric Control Plane
Edge → POP → Region → Control Plane
AUTO MODEON
OP
1

Global Operations Center

Denver POPLondon POPNew YorkSão PauloSingaporeSydney
Global RPS
1.8B
+12.4%
Cache Hit
94.6%
+3.2%
Latency P50
18ms
-4.6ms
Error Rate
0.001%
-0.002%
2

Real-Time Request Flow

Edge1.52B
POP1.18B
Region120M
Live Events● LIVE
EDGE-17 → POP-1HIT
POP-1 → US-CENTRALFETCH
US-CENTRAL → NVMeOK
3

AI Command Center

HIGHPrewarm Denver POP98%
MEDIncrease TTL /video/*92%
LOWAdd Mumbai replica71%
Executed
✓ Shift: 15% US-East14:32
✓ Expand Denver14:30
✓ Failover none14:28
4

Cache Intelligence

Edge Hit
96.1%
POP Hit
88.2%
Hot Objects
1. /video/fifa_final.mp412.4M
2. /images/homepage.jpg8.7M
3. /api/catalog6.1M
4. /css/main.css5.3M
5

Storage Fabric · NomadSDSaaS

US-CENTRAL
● HEALTHY
24ms RTT
EU-WEST
● HEALTHY
27ms RTT
AP-SOUTH
● HEALTHY
31ms RTT
SA-EAST
● HEALTHY
36ms RTT
Total Objects24.6B
NVMe Capacity68%
Replication Lag3ms
Durability11 Nines
6

Incident Command

Denver POP
Elevated NVMe latency · Impact LOW · Since 14:31
AI Root Cause
SSD wear increase on Drive Group 3
Confidence 94%
DrainMigrateReplace
Telemetry & Control Pipeline
Edge Nodes
Collect Metrics
POP Aggregators
Stream Telemetry
Region Mesh
Events & Metrics
Stream Processor
Kafka / Pulsar
AI Decision Engine
Predict → Decide → Act
Control Actions
Route · Cache · Scale
Verify & Learn
Feedback Loop
See everything
Edge, POP, and region telemetry unified into one live console with sub-second freshness.
Decide autonomously
AI Command Center proposes and executes routing, caching, and capacity actions in real time.
Heal in place
Incident Command runs drain → migrate → replace flows without paging a human.
Pricing

Transparent. Flat.

No surprise egress, no per-token AI bills, no hidden tiers. Everything observable, by default.

CDN
$0.01/ GB

Global delivery, AI-routed.

Object Storage
$0.015/ GB / mo

Hot, replicated, no egress fees.

SMTP Email
$0.05/ 1K emails

Edge-routed deliverability.

AI Automation
Included

No tokens. No metering. Ever.

No hidden AI token fees
30–70% lower infra costs
Unified observability included
The Vision

We're building the self-operating infrastructure layer for the next generation of the internet.

A planetary-scale system that routes, replicates, secures, and heals itself — so humans can ship products, not pager rotations.