They can change the model later, add caching, improve performance, and use better hardware. Just getting a native search engine off the ground is impressive.
They may be running a flask app reading data direct from s3 feeding through a giant ML model right now - which wouldn't be the application architecture they'd use if they served 100x the traffic.
They may be running a flask app reading data direct from s3 feeding through a giant ML model right now - which wouldn't be the application architecture they'd use if they served 100x the traffic.