토스 AI/ML 엔지니어의 하루 #토스 #AI #ML — Transcript

Full transcript

1. 0:00Secondly, what kind of environment will
2. 0:01you be working in here? We’ve chosen
3. 0:03a path that doesn't rely on external
4. 0:05parties. We handle our own serving as
5. 0:07well. We’ve built this into a
6. 0:09cloud-based model to handle all serving
7. 0:12, so developers like you can simply
8. 0:14call an API to use any model—whether
9. 0:16commercial or open-source—and consume
10. 0:18as many tokens as you need. Secondly,
11. 0:21we are in the process of turning our
12. 0:23massive amount of data into an ontology
13. 0:25. We aren't just standardizing all data
14. 0:27into marts, but we've reached the final
15. 0:29stages of building semantic layers,
16. 0:31graph databases, and a full ontology.
17. 0:33With around 80 people on our DA and DE
18. 0:35teams dedicated to data maintenance,
19. 0:37we’ve cleared the way for the
20. 0:39application layer. All you need to do
21. 0:41is come in and enjoy yourselves.
22. 0:43We’re building an environment where
23. 0:45you can conduct offline tests in the
24. 0:46afternoon, run a simple pilot
25. 0:48deployment, check the results, and
26. 0:49leave for the day while online tests
27. 0:51run in the background. I want to
28. 0:53provide that kind of free experimental
29. 0:54environment for you. We are creating a
30. 0:56world where the only limits are your
31. 0:58own fears and your own imagination. You
32. 1:00can think of it as us running almost
33. 1:02every AI project currently on the
34. 1:04market, just on a smaller scale.