Saurabh Saxena Employs Real-Time AI for Sentiment Analysis at Uniphore

Saurabh Saxena Employs Real-Time AI for Sentiment Analysis at Uniphore

Saurabh Saxena, Head of Technology and R&D at Uniphore

Video preview
Saurabh Saxena
Saurabh Saxena
Head of Technology and R&D at Uniphore

Saurabh Saxena is focused on creating enterprise software that requires unabashed customer-focus, intuition for the "right", relentless effort, deep-and-wide technology know-how, and can withstand the tests of load and durability. Saurabh is experienced in applied AI, Natural language processing (NLP), Machine Learning, CRM, and large-scale IoT platforms. He has created platforms that for five years running are ranked top in Gartner IoT platforms and have been a founder of multiple technology startups.

Uniphore combines speech AI, NLP, tonal AI and computer vision AI for creating machine-enhanced, delightful end-customer experiences for B2B and B2C businesses. As head of technology at the conversational AI company, Saurabh’s team at Uniphore is combining research and applied AI to create some fascinating applications.

Watch the video to learn how Uniphore leverages real-time data to help organizations improve the efficiency and efficacy of their customer-facing employees.


Uniphore was founded on the basic premise that conversations are the most critical asset that a company owns. We help companies understand their customers more by using AI to augment human understanding of conversation through artificial intelligence, which can give you a lot more information about a human to human conversation than a human can comprehend themselves. My name is Saurabh Saxena. I'm the Head of Technology and the Head of Engineering R&D for Uniphore.

One of the things that happens when you're doing computer vision is you have to literally manage data on almost 200 landmark points on your face in 1/24th second. So now you think,10 participants in a meeting, everyone's expressions being shed or the machine augmenting the human to understand human sentiment at 1/24th of a second clip for the entire period and then look at all the different conversations that you're having across large organizations, that becomes a massive datasets. So that's where Astra plays in.

When we started looking at the data, we were literally looking at maybe about three to four million records being created for a single meeting. And to put that in perspective, when I first ran a test over one hour meeting, it took us 28 hours to process that one meeting. Now we process a one hour meeting in under eight minutes. Knowing that Cassandra was a very stable platform for us and has been for me in the past, we decided to go with Cassandra and that's why we ended up with Astra.

Astra gave us that worry free setup for Cassandra that we wanted to use. Our initial results are showing a very strong correlation with as customer sentiment goes higher, your win percentage goes higher. As customer sentiment goes higher, your deal cycle starts to drop just a tad. So the initial results are very encouraging. By leveraging Astra and not managing a lot of these computes myself, at a very high level we calculated, it's about 2.5 to 3x cost savings, human capital and hardware combined, that Astra allows us to save on.