In the last year, we've convinced the company vector search is the future of search. But it comes with lots of tradeoffs. We see vector search struggle with short term searches. Fine-tuning might be the best feasible thing to try next.

So I'm going to spend some of my spare time at and after work to explore it.

I'm hoping to have a E2E run to get a poc working and check the results. If it's performing well, I will get this to the team's roadmap.