Axiom

You are making AI/ML architecture decisions by committee. Every sprint, your team debates the same three questions — fine-tune or RAG, build the pipeline or buy the wrapper, scale the GPU cluster now or wait. You have tried vendor whitepapers. You have asked the model providers. The answers are always "it depends." You ship without confidence, and the technical debt from those deferred calls compounds every quarter. Axiom is your AI/ML engineering strategy advisor — a persona built to make the calls your team keeps deferring. Not to document options. To make the call and explain the reasoning. Axiom's decision framework was extracted from 200+ real architecture reviews spanning LLM deployment, vector database selection, inference optimization, and ML pipeline design. Every recommendation pattern is grounded in specific production tradeoffs: latency vs. cost at scale, vendor lock-in vs. time-to-ship, fine-tuning ROI at different data volumes. The system knows where the sharp edges are because those sharp edges have already cut. Unlike a generic engineering advisor, Axiom never produces a "here are your options" response without a recommendation — because options without a call are just expensive delay. That rule exists because every architecture decision that stalled in committee eventually shipped late, over-cost, and under-spec. Axiom closes with a position every time. You can push back. But you will not receive a document that defers the hard part back to you. What you get: SOUL.md — Axiom's decision-making identity and reasoning architecture. IDENTITY.md — operational constraints, communication protocol, and escalation thresholds. ML_DECISION_FRAMEWORK.md — the complete build-vs-buy, fine-tune-vs-RAG, and scale-timing decision trees. ARCHITECTURE_REVIEW.md — structured review template with output format and recommendation protocol. Requires Claude Opus or Sonnet via API or OpenClaw.