Use Case Guide

Personalized Recommendations at Scale

Build recommendation systems that understand user preferences deeply. Average what someone liked into a single query vector and let similarity do the ranking — no separate ML pipeline to stand up.

A taste centroid in twenty lines

Average what someone liked, query with the result — runnable as-is on the bundled sample set

import numpy as np
from veep import Client, samples

df = samples.dataframe()          # ~5,000 popular films with plot embeddings

vp = Client.login()
vp.collections.create("movie-recs", tier="hot")
vp.vectors.upsert("movie-recs", dataframe=df)

# Average a few liked films into a "taste centroid," then query with it
WATCH_LIST = ["film-00002", "film-00004", "film-00026"]  # Lion King, Toy Story, Nemo
watched = np.stack([np.array(df[df.id == fid].iloc[0].vector) for fid in WATCH_LIST])
centroid = watched.mean(axis=0)
centroid = centroid / np.linalg.norm(centroid)

results = vp.vectors.query(
    "movie-recs", centroid.tolist(), top_k=5,
    filter={"id": {"$nin": WATCH_LIST}},
)
for r in results:
    m = r.metadata
    print(f"{r.score:.4f}  {m.get('title','?')}  ({m.get('year','?')})")

The full walkthrough — every code block tested against the live service — is the Recommendations from a watch list tutorial.

Recommendation Algorithms

Implement any algorithm with Vector Panda

🤝

Collaborative Filtering

Find patterns in user behavior. "Users who liked X also liked Y" recommendations powered by vector similarity.

  • E-commerce product recommendations
  • Movie and music suggestions
  • Social media content feeds
📊

Content-Based

Match item features to user preferences. Recommend similar items based on characteristics users have shown interest in.

  • News article recommendations
  • Job matching platforms
  • Educational content
🔄

Hybrid Systems

Combine multiple signals for best results. Use collaborative, content, and contextual data together.

  • Netflix-style recommendations
  • Spotify Discover Weekly
  • Amazon product suggestions
📈

Session-Based

Real-time recommendations based on current session. Perfect for anonymous users or immediate context.

  • E-commerce browsing
  • Content discovery
  • Travel planning
🧠

Deep Learning

Use neural network embeddings for complex patterns. Integrate with transformers and other modern architectures.

  • Video recommendations
  • Fashion and style
  • Gaming matchmaking
🎯

Multi-Armed Bandit

Balance exploration and exploitation. Continuously optimize recommendations based on user feedback.

  • Homepage personalization
  • Email campaigns
  • Ad targeting

Measured performance

One collection on the standard Cohere 1M dataset, measured end to end through the public API with VectorDBBench. Real numbers, not projections — methodology and raw results here.

1,620
QPS sustained
Concurrent query throughput held for the full run
0.9652
Recall@100
Accuracy against exact brute-force ground truth
49.5 ms
p99 latency
Slowest 1% of queries at full load
$17.14
That collection / month
Storage-only pricing; the queries were free

Build Recommendations That Convert

Start with Vector Panda and deliver personalized experiences that keep users engaged. Scale from MVP to millions without limits.

Get Started Free →