Cloud cost optimizer
Finance flags the cloud bill every quarter, and engineers rightsizing by hand never keep up. Design an advisor agent that continuously proposes cost optimizations — rightsizing, storage tiers, orphaned resources — but never acts alone. Why is advisory-only the right boundary here, how do its proposals become one-click-safe for engineers, and how do you measure that it's finding real savings rather than quietly breaking headroom? Then make the case to your VP of engineering.
realized_savings_report(statuses: list[str], projected_savings: list[int], realized_savings: list[int], days_to_incident: list[int], incident_window_days: int) → list[int][["applied","proposed","rejected"],[1200,800,500],[900,0,0],[-1,-1,-1],14]out[2500,900,900,33][["applied","applied"],[1000,1000],[1000,950],[3,-1],14]out[2000,1950,950,100][["applied","rejected","applied","proposed"],[300,900,700,250],[250,0,800,0],[-1,-1,-1,-1],14]out[2150,1050,1050,50]Turn the fuzzy goal into a bounded agent task. Say what the agent reads, what it may do on its own versus draft for a human, and what it must never touch. Then earn the trust: how you’d verify it before rollout, and the pitch (the problem, what you built, why it matters).
Agent build: design the agent in plain language (or narrate it). The coach grades the decomposition, the guardrails, the verification plan, and the pitch.