OCS Intelligence LLM โ Phase 2: Deploy, monitor & access
Deploy the shared LLM, watch it under class load, and make sure every OCS student can reach it.
Phase 2 โ Deploy, watch, and make sure the class can get in
Put the service in front of real students, watch it under load, and keep it reachable so access is a class right โ not a lucky connection.
What this phase is for
How the work shows up
Treat the rack as a service the school can point at โ not a one-team experiment that vanishes after a demo.
Notice when answers stall, when memory gets tight, or when the stream drops, before a whole class feels it.
Classroom use is the test. The service should remain usable when several students are mid-assignment together.
Access is the impact. If some students still need a paid tool because they cannot reach OCS Intelligence LLM, the question is unfinished.
Done when
Research question 2 ยท Secondary โ only answerable once RQ1's service exists
Once the service exists, how do we keep it fair, visible, and usable when many students need it at once?
Research endpoint: Answered when ~5 concurrent users hold steady without OOM, a hard crash, or token speed falling to a crawl, and every OCS student โ not just the builders โ can get a key.
Priorities for this phase
If this phase works
Literature
| Resource | Type | What it means |
|---|---|---|
| TTFT, TPS, ITL | Metrics | Time to first token, tokens per second, and inter-token latency โ the numbers we track to know the service still feels fast under load. |
| OCS Intelligence issue #4 | Tracker | GitHub issue that tracks literature, metrics, and serving goals for this project. |