recsys-engineering

Design, implement, debug, and review industrial recommendation-system architecture or algorithms using the released book PDFs, curated examples, and runnable templates.

SKILL.md

GitHub

Recommendation-system engineering

Requirements

Resolve the constraints that change implementation: product objective, QPS, P99 latency, corpus/feature scale, update freshness, consistency, runtime, deployment stack, availability, privacy, and cost. Ask only for materially different tradeoffs.

Evidence

  1. Search the local PDF index for relevant book guidance.
  2. Read cited PDF pages.
  3. Inspect related repository code and templates.
  4. Keep 本书依据, repository convention, and new engineering assumptions separate.

Build

  • Reuse existing interfaces and runnable templates.
  • Keep [MOCK] dependencies clearly labelled and replace them only when real integration details exist.
  • Services require explicit contracts, timeouts, overload/degradation, observability, version consistency, rollback, and capacity assumptions when relevant.
  • Models require objective, sampling, feature-time semantics, offline/online parity, complexity, and evaluation criteria when relevant.
  • Deliver a complete runnable path without placeholders.

Verify

Run the changed surface. Report 本书依据, 工程实现, 验证, and unresolved 待业务确认.