AXIOM was built to run on the most constrained hardware available — a phone, around a full-time job, without a server farm. That constraint proved the architecture. Field and mobile deployment is not an edge case for AXIOM. It is the origin story.
AXIOM's entire architecture was designed and validated on a Samsung S24 Ultra running Termux — no server, no cloud, no dedicated hardware. Every design decision was made under the constraint of limited RAM, limited storage, and no reliable network connection.
That constraint forced better engineering. Epistemic integrity requires architectural discipline, not massive server farms. If it runs correctly on a phone, it runs correctly anywhere.
Field medical teams operating without reliable connectivity query a locally validated clinical knowledge store. When connectivity restores, delta capsules sync safely via CRDT merge — no cold-start storm, no data loss, no conflict.
Defence, intelligence, and secure field contexts where data sovereignty is absolute. AXIOM operates fully air-gapped. No packet leaves the device. The validated store is local. The answer is immediate.
Factory floor, remote site, and industrial environments with intermittent connectivity. Maintenance procedures, safety protocols, and compliance requirements accessible on-device. No dependency on a central server being reachable.
Sales, field service, and inspection teams operating across geographies with inconsistent connectivity. Role-specific validated knowledge on device. Queries answered locally. Updates synced safely when connection is available.
AXIOM's architecture was not designed for constrained hardware as an afterthought. It was proven on constrained hardware from day one. The Justification Engine is deterministic rule-based code — no model inference, minimal compute. The Extractive Semantic Translation Layer is quantised to INT4 for CPU-only execution. The knowledge store is an LSM-tree optimised for sequential writes on flash storage.
Every architectural decision that makes AXIOM viable on a phone makes it more reliable at enterprise scale. The constraint was not a limitation. It was the test. If it runs correctly here, it runs correctly anywhere you need it to.
This architecture was stress-tested using a general LLM as an external review tool. The LLM was not in the pipeline. It never is.