Deep explanation
Design Amazon Bedrock architecture for a coding copilot for a large engineering org
Design a coding copilot on Amazon Bedrock: Converse streaming, embeddings and OpenSearch for a 50M-chunk code index, ACL filtering, IAM and PrivateLink, throughput planning, Guardrails, and evaluation.
Quick answer
I would put a copilot service in the company's AWS account that calls Bedrock through a VPC endpoint, with a fast model for quick completions and a stronger model for chat and multi-file changes, all via the Converse streaming API. A separate indexing pipeline chunks code by function and class, embeds it with a Bedrock embedding model, and stores it in a vector and keyword index such as OpenSearch, with repository ACLs on every chunk. Throughput is planned with quotas, cross-region inference, or provisioned throughput, and quality is measured with acceptance rate and execution-based tests.
The interview question
A large engineering org wants to scale its internal coding copilot from a prototype to production on Amazon Bedrock. It must answer questions and suggest code using context from about 50 million indexed code and documentation chunks, respect repository permissions, and keep source code inside the company's AWS environment. What architecture would you propose, and why?
Deep explanation
Simple explanation
Sign in to unlock the full answer
Free accounts include 5 full deep answers. Sign in to start unlocking.
- 18 more sections of deep explanation
- Real-world examples
- Common mistakes
- Interviewer expectations
- Follow-up questions