The overbuilt AI design
Over budget: $890/mo
- p95 latency (server)
- 46 ms
- target < 150 ms
- Error rate
- 0.0%
- target < 1%
- Est. cost / month
- $890
- budget $300
Challenge · intermediateFix the AI
An AI suggested this for an internal dashboard with 600 requests per second. It works, but it costs far more than the team’s budget. Cut cost while keeping it healthy.
Handle 600 req/s with p95 under 150 ms, errors under 1%, every part under 90% busy, for at most $300/month.
Need a hint?
Check each component’s utilization. Which ones are nearly idle?