Trained a CNN machine learning model using the techniques of transferred learning, which when the user provides it with an image from his catalog predicts the top 5 relevant parts. This eliminated the tedious process for professional from manually find the appropriate parts from Various catalogues. The accuracy of the model is 94.2%.

Trained an NLP ML model to traverse through all the user documents (Installation, help, FAQs, etc.)  and provide the relevant information that the user is looking for with an accuracy of 89.2%. These user inputs and responses are via an azure chatbot, and this chatbot has also been trained to understand the intent of what user is asking.

A Siamese NN has been trained as part of One-Shot Learning to recognize faces with 91.2%. And another Siamese NN for expression recognition with 87.6?curacy, both the networks were used in conjunction to prevent the face spoofing and thus increasing the security.

This system is developed such that there is no need to train the model for every new employee and a single image of an employee is enough to recognize him daily.

Developed merge tool web app which helps in comparison and merging meaningful engineering objects rather than mere database tables. This eliminated the Blackbox approach of upgrading entire databases to selective and modular object-wise upgrades. This is a 2 pane Web-UI tool that looks in similar lines with that of WinMerge

H&S SDB library is the enterprise standard library for creating Hangers and Supports with minimalistic programming. This was achieved by developing predefining generalized classes that house much of the complexity for the creation of supports. User-specific business rules are then fed into the library with the help of extension methods. The library was an instant hit and is used by almost all EPC companies worldwide.