Browse through our catalogue of pre trained models for common text classification and entity
recognition tasks, which can be integrated with your application immediately or after fine tuning.
Ease of Use
Test with your dataset and use it in your application. No need to go through the iterative process
of collecting data, experimenting with models and testing it again and again. You can integrate
these models directly into your application using our SDK to unlock ML capabilities with ease.
Fine tuned to your needs
Using labelled data, our pre trained models can be fine tuned to the task in question for even
better performance, from your browser with all the heavy lifting taking place in our servers.
Trained on Quality Data
These models have been trained on data collected, labelled and annotated in house by our
team to ensure that these models are trained on best possible quality data.
Highly Scalable
Our models are designed to scale with the task in hand. From mobile to private cloud, models
are available for a range of environments.
Leveraging Cutting Edge of ML
These models are built using the the cutting edge of ML techniques such as transformers,
quantized trading etc for the best possible model.
Pre Trained Model to classify text as Spam/Ham. Currently only supported in English. The model has been trained on Text from SMSs, Notification, Emails and can achieve an accuracy of >99. 4% with an f1 score of >0.99
Pre Trained Model to extract OTP from a given text. Currently only supported in English. The model has been trained on Text SMSs and can achieve an classification accuracy of >99.8% with an f1 score of 1 and token identification accuracy of >99%
Pre Trained Model to extract multiple entities. Currently only supported in English The model has been trained on a data set created specifically for this purpose. Achieves >99% accuracy in token identification.
Pre Trained Model to extract multiple entities. Currently only supported in English The model has been trained on a data set created specifically for this purpose. Achieves >99% accuracy in token identification.