About AXE

We build AI in Canada.

Foundation models trained from scratch. Audit-grade deployments on customer iron. Zero external API spend.

Companyfounded|2025headquarters|Toronto, CAteam|4 agents · 2 humansfleet|5 nodesproducts|5 pillarslive|Edge-2 Large2026-05-12
Founders

Two operators. One country. One stack.

Co-founder · CEO

James Lewis

Architect of the AXE stack. Trained the first Edge model from scratch on JL2 (Apple Silicon). Builds in public with the team: Forge, Cortana, Vigil, Ghost. Believes every Canadian organization with serious AI ambitions should own the weights.

Co-founder · COO

Michael Mayo

Operations and procurement. Runs the customer-facing surface that turns the platform into deployments. Brings the discipline that takes Canadian-built infrastructure from prototype to government-grade.

The alternative to building this is sending every Canadian query to a server in Virginia and renting back our own intelligence under someone else's jurisdiction. That isn't ownership. It's rent.

— Founders' note · 2026

What we believe

Five rules. We don't compromise on any of them.

First principles

Trained from scratch on Apple Silicon.

Not a fine-tune of a foreign base. Not a wrapper around a vendor API. The weights, the tokenizer, the data — all built by us, on hardware in Canada.

Zero spend

$0 external API costs.

Local inference on llama-server and Ollama. The customer pays once for hardware; recurring API bills disappear. Ownership is a margin position, not a marketing claim.

Customer-portable

Your iron. Your inference. Your data.

The weights stay AXE-owned · the deployment is yours · the inference happens on your hardware behind your perimeter. Every release-v1 carries a Knox EMA attestation chain so an auditor can replay provenance offline. If we vanish tomorrow, the deployment keeps serving — under your control, with the audit trail intact.

Procurement grade

PIPEDA · Protected B · FIPS 140-2 path.

Knox audit chain · post-quantum signing (ML-DSA-87 FIPS 204 Level 5) · revocation freshness · key rotation. Every claim auditable.

Canadian first

Toronto inference. Not Virginia.

Mountie answers 40 million Canadians via Service Canada integration. Knowing where the query is processed is part of the product, not a footnote.

The fleet

Five callsigns. One Canadian mesh.

NOVAPrimary training cluster · 64 GBFORGECommand + ops · 16 GBVIGILSecurity + eval · 16 GBGHOSTStealth ops · edgeREAPEREdge processing · 8 GB
Where we operate
Toronto-first. Canadian-only customers. Government, defense, and enterprise organizations that require complete data residency — including PIPEDA-aligned health, financial, and classified workloads. We do not deploy outside Canada.
2025
Founded
James Lewis · CEO · Canada
1.34B
Largest Edge model
trained from scratch · converged 2026-05-12
$0
External API spend
local inference only · zero foreign rent
5
Platform pillars
CASTLE · Edge · Algorithm · Knox · Supercharge

Bring your data home.

We deploy on your iron. We integrate with your perimeter. We leave you with the weights.

Request Access