Founded in 1996 by Sridhar Vembu, business software provider Zoho took a leap of faith in artificial intelligence (AI) with the launch of its AI and machine learning (ML) labs in 2014.
Six years later, the once two-member team stands at 100.
“At Zoho Labs, we do very focused research, not just ivory tower research -- research that has a goal to see the light of the day, not research that just goes for publication or conference talks,” director of AI research at Zoho, Ramprakash Ramamoorthy told TechCircle, in an interview.
Having begun with early-stage AI models on databases, which form the core of all enterprise business, the team went on to form a hardware group that worked on accelerating the models and databases using graphics processing units, regular expression computation, used Field Programmable Gate Arrays (FGPAs), among other projects over the years.
Currently, Zoho’s AI and ML business unit functions across three core areas -- statistical machine learning, computer vision, and natural language processing (NLP).
“We used to joke internally that statistical machine learning is a statistics group with a marketing team. Things have been around since the 1950s-60s. But then it's good, you train a model on top of your data and then you take intuitions out of it. Typically we have something like a time series analysis, you do apply anomaly and forecasting… again, these are much better versions than their non-AI and ML counterparts. AI and ML account for multi-dimensional values where trends and seasonality are captured,” Ramamoorthy said, when asked about use cases that have been developed across the three core areas of work.
In the computer vision work area, he said that currently, images and data form a large chunk of enterprise software. “Let's say I go on a business trip. ‘Zoho Expense’ can take photos of all the recipes, carry out Optical Character Recognition (OCR), forecasting, and anomaly algorithms to simplify the whole princess with the OCR suite,” he explained.
Zoho Notes as an application is also a use case from the unit’s NLP work area. When users visit conferences and take pictures on their phone, it lets them turn the images into a note format, which are also searchable through images. Data interaction in NLP happens via texts, translation, grammar, and language.
In the video, which are excerpts from the interview, Ramamoorthy also talks about early days of AI in enterprise businesses in 2011, recurrent neural networks and computer vision as right fits into Zoho's existing products, the return of investment of AI research at Zoho, and how the technology will change future business software.
Ramamoorthy graduated with an undergraduate degree in information and communication technology engineering in 2012, and began his journey with Zoho as an intern, back in 2011.