Developing generative AI solutions
Course Summary
The whole delivery business
The story: Looking back at the delivery business: you know how to pick a driver, how a map book helps, what each kind of training costs, why a dispatcher matters for complicated days, and how to test drivers properly. Knowing the whole process is what makes the business reliable.
In AI/AWS terms: Knowing the whole generative AI lifecycle, from use case to deployment, is what makes a generative AI system robust and trustworthy. The course covered:
- Criteria for choosing a pre-trained model
- RAG and its business applications
- Cost trade-offs between prompt engineering, RAG, fine-tuning, and building from scratch
- Agents for multi-step tasks
- Evaluation methods and metrics for FMs
For the exam: Remember the cost order (prompt engineering < RAG < fine-tuning < from scratch) and which metric goes with which task.