Beginner-Intermediate

Use Cases & Case Studies

When to Use Qdrant

Retrieval-Augmented Generation (RAG)

Store document chunk embeddings and retrieve the most relevant ones when users ask questions. Qdrant's payload filtering enables scoping retrieval by document category, date range, or access level -- critical for production RAG with multi-tenancy.

Semantic Search

Search that understands meaning rather than just keywords. "Affordable family vacation" returns results about "budget-friendly trips with kids" even though no words overlap. Hybrid search (dense + sparse vectors with RRF) delivers the best of both worlds.

Recommendation Systems

Embed items and users into the same vector space, then find items most similar to a user's preference vector. Qdrant's recommendation API supports positive and negative examples for nuanced "more like this, less like that" suggestions.

AI Agent Memory

AI agents store embeddings of past interactions and learned facts. When recalling context, they query Qdrant with the current conversation embedding. Payload filtering enables time-based and topic-based retrieval.

Anomaly Detection

Embed normal behavior patterns as vectors. New observations far from all stored vectors (low similarity) are flagged as anomalies. Qdrant's fast nearest-neighbor search makes this real-time.

When NOT to Use Qdrant

Not a general-purpose database. No SQL JOINs, GROUP BY aggregations, or ACID transactions. Use PostgreSQL for relational workloads.
Small datasets (< 10K vectors). HNSW overhead isn't justified. Use brute-force search in NumPy or qdrant-client local mode.
Write-heavy transactional workloads. Writes go through WAL and background optimization. Not designed for high-frequency, low-latency writes with immediate consistency.

Real-World Deployments

Tripadvisor
Travel & Hospitality
Powers semantic search and recommendation across millions of hotels, restaurants, and attractions. Vector search understands travel intent ("cozy beachfront with snorkeling") while payload filtering handles structured constraints (location, price, rating).
HubSpot
CRM & Marketing
AI-powered CRM features: finding similar companies, matching contacts to ideal customer profiles, and content recommendations. Multi-tenancy via payload-based isolation serves their massive customer base securely.
Bazaarvoice
Product Reviews & Commerce
Processes billions of product reviews across thousands of brands. Semantic search finds similar complaints across brands and matches descriptions to relevant reviews. Horizontal scaling handles data volume.
Bosch
Manufacturing & IoT
Anomaly detection in manufacturing: sensor readings are embedded and compared against historical patterns. Engineers also search millions of technical documents using natural language queries.
OpenTable
Restaurant & Dining
Improved restaurant recommendations by embedding diner preferences and restaurant characteristics. Considers ambiance, price sensitivity, and occasion type beyond simple cuisine matching.

📚 References & Resources