01 · Routing
Intelligent Routing
Local models handle simple tasks; complex queries auto-route to cloud LLMs — optimizing cost and latency in real time.
PRIVATE BETA
·
SDK ALPHA · Q2 2026
AI on Every Device.
Even a $10 One.
Netralis is the OS layer that brings large language models to low-cost microcontrollers — voice assistants, AI companions, and smart controllers on hardware anyone can afford.
// Three lines to a full voice AI assistant
#include "netralis/agent.h"
auto agent = NeAgent("home-assistant");
agent.on_wake("hey netralis", []() {
// STT → cloud LLM → TTS, all managed by Netralis
auto query = mic_listen(); // streaming ASR
auto reply = agent.ask(query); // auto-route: local or cloud
speaker_play(reply);
});
netralis_start(); // OTA, routing, multi-lang — batteries included
#include "netralis/agent.h"
auto agent = NeAgent("home-assistant");
agent.on_wake("hey netralis", []() {
// STT → cloud LLM → TTS, all managed by Netralis
auto query = mic_listen(); // streaming ASR
auto reply = agent.ask(query); // auto-route: local or cloud
speaker_play(reply);
});
netralis_start(); // OTA, routing, multi-lang — batteries included
Cost
<$10
Target hardware
Platform
3 chips
ESP32-S3 · C3 · P4
Founder
15 yr
Engineering experience
Scale
1B+
Users on platforms shipped
Platform Capabilities
Everything the intelligent
edge needs
Four core layers — from hardware abstraction to cloud LLM routing — so you ship the experience, not the plumbing.
02 · Voice
Voice-First Interface
Streaming ASR + LLM + TTS with wake-word, speaker recognition, and multi-language support out of the box.
03 · Hardware
Hardware Abstraction
One OS, many chips. Unified APIs for sensors, displays, audio, and OTA across ESP32-S3, C3, and P4.
04 · Agent Runtime
Agent Runtime — MCP Protocol
MCP-based agent framework lets third-party developers build AI apps that run on any Netralis-powered device. No firmware expertise required — just write your agent logic and deploy over-the-air.
Cloud Infrastructure
Built on Google Cloud
Netralis's cloud layer runs end-to-end on Google Cloud — from large-model inference to device fleet telemetry — so we focus on the edge while Google handles the planet-scale plumbing.
Vertex AI · Gemini
Cloud LLM Inference
Gemini Flash handles complex queries that exceed on-device model capacity. Streaming responses keep latency under 800 ms for voice.
Pub/Sub · IoT Core
Device Fleet Messaging
Bidirectional messaging for OTA updates, agent deployments, and remote configuration across millions of low-power microcontrollers.
Firebase Auth
Identity & Provisioning
Developer accounts, device identity, and end-user pairing — with secure enrollment flows for both consumer and enterprise hardware.
BigQuery
Telemetry & Routing
Aggregated, privacy-preserving usage signals feed back into the routing layer to continuously improve local-vs-cloud decisions.
Roadmap
What's shipping when
Concrete, dated milestones from private SDK to commercial release. Updated as we ship.
Q1 2026 · Done
Architecture & Public API
Hardware abstraction layer, routing protocol, and MCP runtime architecture defined. Public SDK API drafted and published on GitHub.
Q2 2026 · In Progress
Private SDK Alpha
ESP32-S3 baseline, voice pipeline (ASR + LLM + TTS), MCP runtime stub, OTA updates over Pub/Sub.
Q3 2026 · Planned
Developer Dev Kit
Reference hardware kit, onboarding flow, and open enrollment for the first 50 external developers.
Q4 2026 · Planned
Public Beta
Cross-chip support (ESP32-C3 + P4), MCP agent marketplace, public docs, and open waitlist conversion.
Founder
Built by a founder who's
done it at scale
FOUNDER · ACTIVE
George Guo
Founder & Engineer · Netralis Inc.
15 years of software engineering experience, including a decade at Ant Group (Alipay) — one of the world's largest fintech platforms, serving over 1 billion users and processing trillions of transactions annually. Now based in Edmonton, Canada, building Netralis to make LLM intelligence accessible on the cheapest hardware on the planet.
- 10 years at Ant Group / Alipay — distributed systems and payments infrastructure at global scale
- 15 years of Java, cloud architecture, and large-scale system design
- Based in Edmonton, Alberta, Canada 🇨🇦
Early Access
Join the waitlist
Building on ESP32 hardware? Investing in edge AI? Drop your email — we'll reach out when the SDK alpha opens in Q2 2026.
✓ THANKS · GEORGE@NETRALIS.CA COPIED TO YOUR CLIPBOARD. EMAIL US TO CONFIRM YOUR SPOT.
OR EMAIL DIRECTLY · GEORGE@NETRALIS.CA