feature-store plugins

by Matt Corbett

ML-engineering (MLOps) team — agents (ml-platform-architect, training-pipeline-engineer, model-serving-engineer, ml-monitoring-engineer) for the PRODUCTION lifecycle of ML models: the platform/architecture (build-vs-buy, the stack, the train->register->serve->monitor loop), reproducible training pipelines + experiment tracking + a model registry, feature stores and train/serve consistency (avoiding training-serving skew and leakage), model serving (online vs batch, shadow/canary), monitoring (data + concept drift, decay, retraining triggers), and computer-vision MLOps (task->architecture + edge-vs-cloud inference). skills, a decision-tree knowledge bank (serving-pattern + retraining + computer-vision trees + a dated 2026 map), best-practices, templates, commands, an advisory hook. Seams: significance -> applied-statistics, data pipelines -> data-platform/data-streaming, LLM/agent apps -> claude-app-engineering, deploy -> devops-cicd/cloud-native-kubernetes. Requires ravenclaude-core@>=0.7.0.

by Matt Corbett

Recommendation-systems-engineering team — the RECSYS layer that turns interaction data into ranked recommendations. agents (recsys-architect, recsys-implementation-engineer) covering approach selection (popularity baseline, collaborative filtering, content-based, hybrid, two-tower/embedding retrieval, sequential), the candidate-generation → ranking → re-ranking pipeline, cold-start (new user/new item), offline evaluation (recall@k, nDCG, MAP, coverage/diversity) AND online A/B, feature stores with train/serve parity, low-latency serving, and feedback-loop/position bias. skills, Mermaid-backed knowledge docs, templates. Seams: training infra → ml-engineering; keyword/semantic search → search-relevance-engineering; A/B + stats → experimentation-growth-engineering + applied-statistics. House line: baseline before a neural net; offline wins must survive an online A/B. Requires ravenclaude-core@>=0.7.0.