@ibocodes LLM Judge usecases can pretty much be replaced with Jev models at dirt cheap cost and latency. Fast, narrow classifications without reasoning are best suited.
Consider using pagination here, don’t ever return 1M rows!!
Prefer cursor/keyset pagination over offset:
GET /users?limit=100&cursor=<last_id_or_created_at>
Query: WHERE (created_at, id) > (?, ?) ORDER BY created_at, id LIMIT 100
We also need to add the right indexes on the cursor fields (e.g., (created_at, id)).
Also return only the needed fields (projection), and compress response (gzip).
If we are talking about huge exports, then we need to make it async:
start job → return jobId → download from S3.
System Design Playbook:
Giveaway Alert!!
• System design fundamentals.
• Condensed notes to read before the system design interview.
• Must know concepts from real-world software engineering case studies.
(24 hours only!)
To get it:
1. Like, Retweet & Follow @systemdesignone
2. Reply "Playbook"
Then I'll DM you the details.