Who are we
Founded in 2018 in Poland, DEVOPSBAY is a team of 60 engineers expert in machine learning, software development, container environments and Kubernetes applications who have successfully launched 100s of digital products with world leading AI companies.
emerged from a DEVOPSBAY research project, which became an independent spin-off, exploring different applications of image recognition and processing of our proprietary AI models. We are a small self-founded team, based in US, Poland and France, with 10 full-time staff and an incredible set of advisors.
What drives us
We are convinced that AI has a wide range of applications ahead that will have a significant impact in the real world of business, that will empower people, help to streamline work processes and give consumers more information to make informed decisions.
Our unique AI technology
is a cutting-edge AI technologies that identifies and categorizes objects within images, by combining robust visual search, precise object segmentation, a fusion of localized and personalized Large Language Models (LLMs), and trained with expert datasets. We developed this innovative architecture that exploits the power of models, combined with our proprietary object classification model:
- Object segmentation - customized, enhanced version based on Segment Anything
- Object classification - using our own trained proprietary model (trained with our own datasets for results relevant to e-commerce)
- Brand/logo detector - using our own trained proprietary model
- Optional character recognition (OCR) - customized, enhanced version based on Keras-OCR
- Product description generator - customized, enhanced version based on GPT-J NEO 6B and newer version LAMA
Experiments have demonstrated greater accuracy to identify products than the best-in-class current solutions, such as Google Lens. In addition, we have added advanced Web Crawlers that feeds back up-to-date data to LLMs that highly improve content generation accuracy, compared to other solutions such as ChatGPT.
Provides a holistic view of the product from both visual and informational perspectives, setting us apart in terms of accuracy and innovation
How it works
d
e
s
c
r
b
Load a photo
Customize and set your requirements such as desired language style, descriptions length, etc.
Image recognition
segmentation, feature extraction and classification to identify product, brand, logo, color; Deep Learning, OCR for text
Contextual understanding
Beyond just identifying an object, the system will be able to detect all products worn by a person, and to understand the context
Advance learning
Search identical or similar products on the web using advanced LLM based web crawlers to feed information into our LLM
Description generation
Unique product description generation - seo compliant, translated in all languages
Done!
You can: customize, validate, export to templates or directly via api to your vendors
Continuous learning
Your changes will be taken into account to provide continuous improved results
user
Load a photo
Customize and set your requirements such as desired language style, descriptions length, etc.
Done!
You can: customize, validate, export to templates or directly via api to your vendors
d
e
s
c
r
b
Image recognition
segmentation, feature extraction and classification to identify product, brand, logo, color; Deep Learning, OCR for text
Contextual understanding
Beyond just identifying an object, the system will be able to detect all products worn by a person, and to understand the context
Advance learning
Search identical or similar products on the web using advanced LLM based web crawlers to feed information into our LLM
Description generation
Unique product description generation - seo compliant, translated in all languages
Continuous learning
Your changes will be taken into account to provide continuous improved results
All in seconds!