Ecommerce Embeddings for better shopping search

Introducing Mighty Ecommerce - the best embedding model for your store's site-search.

Enables customers to find the products they're looking for more quickly, leading to higher satisfaction rates and increased sales.

High Precision and Recall. Outperforms other embedding models by a wide margin.

Really Fast, no GPU needed! Up to 180 embeddings per second on a single CPU core.

Flexible experience.Designed for Mobile, Tablet, and Desktop result pages.

Deploy in your stack with zero lock-in. Build your store on your terms, not someone else's expensive API.

Easily plug it in to your existing site-search with Mighty connectors!


Latency and Throughput Profile

Example for 10,000 query requests, using only one CPU core.

Add as many cores as needed for linear scalability!

Latency: Returns vectors in 8 milliseconds

Throughput: 136 queries per second

Summary:
  Total:  73.2846 secs
  Slowest:  0.0119 secs
  Fastest:  0.0071 secs
  Average:  0.0073 secs
  Requests/sec: 136.4544
  
  Total data: 619100001 bytes
  Size/request: 61910 bytes

Response time histogram:
  0.007 [1]   |
  0.008 [9358]|■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■
  0.008 [637] |■■■
  0.009 [2]   |
  0.009 [0]   |
  0.010 [0]   |
  0.010 [0]   |
  0.010 [0]   |
  0.011 [0]   |
  0.011 [1]   |
  0.012 [1]   |


Latency distribution:
  10% in 0.0073 secs
  25% in 0.0073 secs
  50% in 0.0073 secs
  75% in 0.0073 secs
  90% in 0.0074 secs
  95% in 0.0076 secs
  99% in 0.0078 secs

Details (average, fastest, slowest):
  DNS+dialup: 0.0000 secs, 0.0071 secs, 0.0119 secs
  DNS-lookup: 0.0000 secs, 0.0000 secs, 0.0005 secs
  req write:  0.0000 secs, 0.0000 secs, 0.0011 secs
  resp wait:  0.0073 secs, 0.0071 secs, 0.0109 secs
  resp read:  0.0000 secs, 0.0000 secs, 0.0007 secs

Status code distribution:
  [200] 10000 responses
            

Connects to any Vector Search Engine!

Accuracy Results*

Model ndcg@1 ndcg@4
max.io/ecommerce-encoder-v01 (Ours) 0.77 0.73
sentence-transformers/all-MiniLM-L6-v2 0.68 0.65

*nDCG for Amazon-Science ESCI ranking benchmark.

Bundle it with a Mighty cross-encoder for even better results!