1
Curious builder0 XP earned · 300 to level 2
0 daysFinish a lesson to begin
Badge collection0 of 6 unlocked
51 small wins to finish your pathNext question →
How do you choose between building and buying each layer of the LLMOps stack?
30-second answerSay your answer out loud first, then reveal.
Typical starting point (varies by company)
| Layer | Common choice | Notes |
|---|---|---|
| Model access | Provider APIs + gateway (open source or managed) | Keep provider-agnostic |
| Tracing / observability | Managed LLM observability or OpenTelemetry into existing APM | Ensure export of traces |
| Evals | Open-source frameworks + custom judges; platform tools as scale grows | Datasets are your asset; keep them portable |
| Prompt management | Git-based or platform registry | Must integrate with CI and flags |
| Vector DB | Existing database extension (pgvector) or managed vector DB | Depends on scale and filtering needs |
| Inference serving | Managed endpoints, or vLLM/SGLang on Kubernetes | Self-host only with volume or control needs |
| Guardrails | Classifiers + policy code; frameworks where they fit | Policy logic is usually custom |
| Orchestration | Framework (LangGraph, etc.) or thin custom code | Avoid deep lock-in to opaque abstractions |
| Fine-tuning | Managed fine-tuning APIs or open-source trainers | Depends on data control needs |
Evaluation criteria per tool: security and compliance (SOC 2, data residency, self-host option), interoperability (OpenTelemetry, OpenAI-compatible APIs, MCP), pricing at scale, vendor viability, and exit cost.
Related
This is what real progress feels like.