Skip to content
Permalink
main
Switch branches/tags

Name already in use

A tag already exists with the provided branch name. Many Git commands accept both tag and branch names, so creating this branch may cause unexpected behavior. Are you sure you want to create this branch?

ocean-he/use-case.md

Go to file
 
 
Cannot retrieve contributors at this time
DeepOpinion - FHE use-case Goal Constraint Solution
12 lines (8 sloc) 1.97 KB
Raw Blame
Open in GitHub Desktop

DeepOpinion - FHE use-case

Goal

A leading utility company powering one of the globally most vibrant cities is continuously improving and digitizing its customer service. They receive hundreds of messages on a daily basis from customers including requests, inquiries, and complaints. The customer service team partnered with DeepOpinion to deploy a cognitive automation solution for automating customer tickets analysis and routing to the responsible team for addressing such requests. The data come from multiple systems including customer satisfaction surveys, customer complaints management tools, and live chatbots.

Constraint

The company places high emphasis on customer data privacy and security as the data include sensitive information, such as addresses, contact numbers, names, and more. They are also legally obliged to not send out or process customer data outside the country.

Solution

DeepOpinion developed a machine learning-based cognitive automation solution that fulfilled all the requirements for automating tickets analysis and routing. To additionally fulfill the aforementioned constraint, the IT security team decided to work with on-premise hosting of a solution to protect customer data privacy and ensure legal compliance.

While DeepOpinion successfully delivered the platform on-premise, this is likely not a sustainable approach for cutting-edge ML solutions. On one hand, on-premise solutions make maintenance, upgrades, selling of new products, and services more challenging. On the other hand, DeepOpinion must also release its model to the customer. An ideal approach would be a more secure data exchange over the cloud in a way that meets legislative requirements enforced on such companies. For example, if it is possible to train a machine-learning model that can directly operate on encrypted data, the customer’s data privacy is ensured on one hand, and on the other hand, DeepOpinion can still keep their own machine learning model private.