InferQuest
← Quest map
Inference Phase 6 · Proof of Work

Open Sourcerer

Merged PRs into the engines everyone runs.

0/4 · 1050 XP
You’re previewing the curriculum. Sign in (top right) to track progress, take knowledge checks, and unlock the verifiers.
Briefing

The repo allowlist doubles as the hiring-signal list, with vLLM above all: Red Hat, which employs much of its core team, literally lists contribution familiarity as a plus. SGLang and FlashInfer count equally. The request trace you wrote in vLLM, Deeply is your map here; the best first issues are the ones you can locate in that trace within minutes. Expect the social half (CI, review latency, maintainer taste) to be half the work, and budget patience for it. That isn't friction around the skill. It is the skill.

The three verified PRs escalate on purpose. The first proves you can land anything at all. The second, performance or correctness work with benchmark numbers in the description, proves you can do the actual job. The third turns a data point into a pattern, which is exactly what a hiring manager scanning your GitHub is looking for.

Tasks
  • vLLM is the highest-signal repo (Red Hat literally lists contribution familiarity as a hiring plus). SGLang and FlashInfer count equally here.

    oss+100 XPresource ↗
  • Checked live against the GitHub API: exists, merged, non-trivial.

    oss+300 XPauto-verified
  • One with benchmark numbers in the description. Same repo list, same live verification.

    oss+350 XPauto-verified
  • Three merged PRs is a pattern, not a fluke — that's what a hiring manager sees.

    oss+300 XPauto-verified