On this page

People ask for my exact stack a lot. The answer is less about specific products and more about operational fit.

🔗What I care about when choosing tools

  • reliability under real workloads
  • clear observability hooks
  • reasonable operational complexity
  • cost transparency

🔗Typical stack categories

  • Model providers: selected based on workload characteristics and cost/latency tradeoffs
  • Orchestration: routing, retries, guardrails, and consistent interfaces
  • Data layer: transactional + vector search where retrieval is needed
  • Telemetry: logs, traces, and usage/cost metrics from day one
  • Delivery layer: simple deployment paths the team can actually support

🔗What I avoid

  • shiny tooling with weak operational support
  • brittle custom frameworks before validation
  • stacks that only one engineer can maintain

🔗Rule I keep coming back to

Use the minimum stack that can hit your reliability and ROI goals.

Complexity is a tax. Pay it only when there’s a measurable return.

🔗Next step

If you're deciding whether this approach fits your business, start here: