Back to problems

Design an AWS fine-tuning platform for LLMs

System Design · OpenAI · Hard

Design a managed, multi-tenant AWS service that enables customers to fine-tune their own large language models. The platform must support the following user capabilities: Upload training datasets containing free-form text and optional paired instructions/responses. Select a foundation model and tuning technique, including full fine-tuning, parameter-efficient LoRA/QLoRA, or adapter-based approaches. Submit and monitor training jobs, inspect progress, and evaluate the…

Checking your access…