Database performanceVerified customer engagement
From two full-time tuning experts to 10-minute reviews
- Environment
- SQL Server and AWS Aurora MySQL — around 30 databases holding ~15 TB.
- Challenge
- After migrating from older versions and continuously adding features, the databases developed unexplained slowdowns.
- AI-automated
- Collection, analysis and remediation proposals are all handled by AI.
- Human review
- People only request work and review the AI’s proposals.
ResultBefore: two tuning specialists on-site (~16 h per case). After: a specialist spends ~10 minutes reviewing the AI’s proposal before delivering it to the end user.Cost optimizationVerified customer engagement
Monthly infrastructure spend cut by 60%
- Environment
- Hybrid on-premise + cloud infrastructure relying on high-spec database servers.
- Challenge
- Over-provisioned high-spec servers drove monthly infrastructure cost to ¥20M.
- AI-automated
- Redistributed and right-sized workloads across on-premise and cloud.
- Human review
- Change scope and cost thresholds approved by the customer.
Result¥20M → ¥8M per month — a ¥12M monthly saving, measured against the monthly infrastructure invoice.Development → ProductionVerified customer engagement
From a mockup to a live service in one week
- Environment
- Azure. The customer hands a Claude-built mockup to the GIIP FDE Box.
- Challenge
- Turn a mockup into a running end-user service — build, infrastructure, database and domain — without a standing engineering team.
- AI-automated
- FDE Box builds in Dev, provisions infrastructure via az CLI, sets up the database and domain mapping, and serves end users — then self-checks in operation.
- Human review
- The customer drives requests; the FDE Box responds to them.
ResultDelivered in one week and launched normally. Incidents are handled automatically and on customer request.
