Vody has built models for e-commerce leaders such as Wayfair and Nike.
Vody is focused on multimodal models that can greatly outperform text-only systems.
Vody's models can be quickly fine-tuned on any product database.
Our models can be deployed in easy-to-use cloud environments or entirely on your hardware.
Unlike general purpose models, Vody purpose builds and domain adapts to improve retail product search and product recommendations.
We've built 14 models that perform tasks across classification, generative, and embedding creation.
Classification: 28 images/second thoughput. 500M parameters. color: 329. pattern is ~50. integrated object detection makes it faster/simpler/enables zero-shot. E2E.
NER model: extracts name with explicit labeling.
Style model furniture: understands midcentury modern, coastal, industrial, country/farmhouse.
Fashion -specific model: button-down, crew-neck
Lifestage: teen, baby, adult.
Material & finish: Home Depot
Room: Walmart.
Embeddings: takes product title/dsec/spec/image. unlimited dataset size. can adjust threshold for confidence. 768-digit vector. fast cosine similarity algo to generate similaity score. or custom algo if needed.
Generative: 4096 token input. billions of parameters. (smaller for faster). can run on A10 instead of A100s. finetuned for ecommerce applications.
Lorem ipsum dolor sit amet, consectetur adipiscing elit, sed do eiusmod tempor incididunt ut labore et dolore magna aliqua. Ut enim ad minim veniam, quis nostrud exercitation ullamco laboris nisi ut aliquip ex ea commodo consequat.
Enterprises face many challenges that prevent them from rapidly maturing their ML Ops and deploying models into production. With Vody they can quickly deploy solutions that: