Model types and performance bottlenecks | Google Cloud Tech | Podwise