Open
Description
ML Recommendation systems don't actually work/ work well unless we have a very large dataset of users and items (jobs, in this case).
Hence it doesn't make sense to use the system itself during launch.
Here's what we need before launch:
- Tracking user activities such as views, impressions, bookmarks, applications to jobs
- Noting down user feed preferences upfront when they sign up (like twitter)
once this data collection infrastructure is in place, we can easily test the recommendation engine while building it and roll out to users.
Creating a recommender system is part-art, part-science, and we can only ever start testing them online, when we have the data.
first priority is the data collection itself for this system in the future.
I'll try to build the initial recommendation engine also. But without data, there would be no way to test it!