Top News
Next Story
NewsPoint

How Hike's VP- AI is leveraging ML to create a personalised UX

Send Push
ET CIO
12th September, 2019 07:00 IST
“We are continuously building in the field of AI & ML to provide a better experience to our users so that they are able to converse seamlessly. Fueled by a culture of research-led innovation, data-driven insights and real-time application of industry-first practices, we’ve used AI & ML at the heart of our product in unique ways,” says Ankur Narang, VP, AI and Data Technologies at Hike.

What has been Hike’s approach towards AI and ML to drive digital innovation?

Our research areas include NLP, computer vision & social network analysis. Under NLP we are creating contextual experiences by making sense of chat in different Indian languages. Through computer vision, we are reinventing audio & visual communication. With social network analysis, we are mining large scale social networks for connecting people & driving business growth.

Hike has been applying machine learning to provide better and more personalized and contextualized experiences for its users.

For hyperlocal engagement between users, we are currently working on some unique elements in NLP to develop unique solutions that can make sense of the text in Indian languages written in the English script. This is where we see AI & ML at the heart of the product, enabling scale and making the overall experience seamless.

Hike has developed ML-backed stickers which offer suggestions for the right sticker during a chat while a user types. The sticker recommendations are delivered through a combination of various parameters — Message prediction and Sticker mapping-- which reduces time and effort for the user to pick the right sticker at the right time.

How are you using analytics in customer segmentation and hyper-personalization of offerings?

Analytics plays a key role across the platform. Some of these include:

• Behavioral analysis per customer/ segment using pattern analysis algorithms

• Personalization signals extraction: gender, nature of relationships, time of day, context of discussion

• Deep Learning & Reinforcement Learning algorithms to predict and score the right stickers at real-time

• Hyper-local and generic context-aware tuning of the algorithms to be relevant and culturally appealing to users.

How do you ensure data anonymization and security of customer data?

In order to deliver on our hyper-personalized offerings, we only need the usage pattern of the user. We don’t need their identity or any of the private data. For this, we remove all private data of the user like mobile number, credit or debit number and other details, and completely anonymize the identity of the users.

All our models are built over the frequent messages exchanged between a community. We only have access to restricted data samples which we use for ML algorithms validation. The identity and all personal data of the user is anonymized and secure in the process since the entire information flow is encrypted.

Moreover, since user privacy is of the utmost importance to us, all user data is safe and secure on the platform. We recently upgraded our security overall and we’re just a step away from military-grade security. Here are 3 things we’ve done: 128-bit AES & 2048-bit RSA Encryption, transport layer security for data ‘in transit’ and ‘at rest’, and SSL pinning for safer communication.

Explore more on Newspoint
Loving Newspoint? Download the app now