The hard part is not the retrieval pipeline, it is the boundary. Embeddings and the index live inside the VPC; the only question is whether the generation model is an in-VPC open model or an external API with a zero-retention DPA, and what crosses that line.
Updated Aug 2026 · Grounded in real Forward Deployed Engineer interview loops and written to a senior-engineer editorial bar.
The hard part is not the retrieval pipeline, it is the boundary. Embeddings and the index live inside the VPC; the only question is whether the generation model is an in-VPC open model or an external API with a zero-retention DPA, and what crosses that line.
The senior signal is naming exactly what leaves the boundary (the assembled prompt plus retrieved snippets, nothing else) and treating that as the thing to redact and audit. Strong candidates separate the two real options (in-VPC open-weight model vs external frontier API under a zero-retention agreement) and pick based on the customer's actual rule (data-cannot-leave vs we-need-a-DPA). Watch for candidates who send raw documents to the model, or who think VPC peering to a SaaS endpoint counts as in-VPC.
No comments yet — be the first to share your approach.
