chief-ai-officer-advisor

Run model_buildvsbuy_calculator.py with your use case profile covering volume, latency, accuracy, and team capacity. It compares 3-year TCO across API tiers, fine-tuning, and self-hosting, then recommends a path with failure modes. Default to API unless you have domain-specific behavior, labeled data, and ML capacity.

Run ai_risk_classifier.py with a JSON profile describing the domain, deployment geography, automation level, and decision impact. It returns a tier (prohibited, high, limited, or minimal), required obligations with article citations, and whether conformity assessment is needed.

For 70B-class models on rented GPUs, breakeven against frontier-economy APIs typically falls between 100M and 500M tokens per month depending on utilization and GPU rates. Run ai_cost_economics.py with your workload to compute the exact crossover with low, mid, and high GPU pricing scenarios.

No. This skill is strategic only and explicitly excludes tactical AI/ML engineering. For RAG implementation, agent design, prompt governance, eval infrastructure, or inference cost optimization, it directs you to the corresponding engineering skills in the repository.

Hire an AI engineer first at Series A to own prompts, evals, and deployment. Research scientists are only justified at Series C or later when the model itself is core IP, since earlier hires lack the eval infrastructure and validated use cases to deliver value.

No. The classifier is triage tooling that surfaces risk tiers, triggered laws, and required controls with citations. EU AI Act conformity assessments and binding compliance decisions require qualified AI counsel, and the skill states this limitation explicitly.