Can I run experiments with different recommendation models?

Yes, the skill includes a Model Serving component and an A/B testing framework that supports deterministic variant assignment and significance testing.

Is this skill compatible with modern Python stacks?

Yes, the provided implementation patterns use FastAPI, Redis-py, and Prometheus-client, making it ideal for modern Python-based microservices.

How does this skill handle high latency in recommendations?

It implements a tiered caching strategy using L1 in-memory storage and L2 Redis caching to serve recommendations with sub-millisecond response times.

What metrics are tracked by the monitoring system?

It integrates with Prometheus to monitor Click-Through Rate (CTR), conversion rates, P95 latency, catalog coverage, and recommendation diversity.

Does it address the cold start problem for new users?

Yes, the skill provides implementation patterns for popularity-based fallbacks to ensure a good user experience when historical data is unavailable.

Recommendation System Engineering

Name: Recommendation System Engineering
Author: secondsky

bysecondsky

•

Data Science & ML

Deploys production-ready recommendation architectures featuring multi-tier caching, feature stores, and automated A/B testing frameworks.

This skill provides a comprehensive blueprint for building and scaling high-performance recommendation engines within the Claude Code environment. It bridges the gap between raw machine learning models and production APIs by implementing essential infrastructure components like Redis-backed feature stores, L1/L2 caching strategies, and robust monitoring for CTR and diversity metrics. Whether you are addressing cold-start problems, thundering herd issues, or complex A/B testing scenarios, this skill offers production-grade patterns to ensure sub-millisecond latency and high-quality user personalization for e-commerce and content platforms.

Key Features

0111 GitHub stars

02Feature store implementation for real-time user and item attribute management

03A/B testing framework with Thompson sampling and statistical significance logic

04Quality monitoring for CTR, conversion, diversity, and catalog coverage

05Automated handling of common issues like cold starts and cache invalidation

06Multi-tier caching (L1/Redis) for sub-millisecond recommendation latency

Use Cases

01Scaling personalization APIs for high-traffic e-commerce or content platforms

02Optimizing backend performance and reducing database load through advanced caching

03Implementing experimental frameworks to test and validate new recommendation algorithms

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add secondsky/claude-skills recommendation-system

For use in Claude.ai and ChatGPT

Download Skill