In India a new data revolution led by cheap Internet tariffs and expulsive growth in consumer Internet platforms are giving rise to rich data sets, which in turn can be used for consumer analytics and retargeting. Increasingly small to large enterprises and tech startups have begun turning towards cloud computing resources to stores, process, and eventually perform analytics to better serve their customers.
In an interview, Navdeep Manaktala, director and head of Digital Native (Startup) Business, Amazon Web Services (AWS) India, on the sidelines of AWS’ annual developer conference in Las Vegas recently, spoke about how new and emerging data sets are persuading Indian startups and enterprises to migrate to the cloud. Edited excerpts:
How are B2B startups approaching the cloud? Do B2B startups host and operate differently on the cloud compared to direct to consumer startups?
B2B companies have done tremendous amount of groundwork to build businesses which are as large in scale as the B2C startups. So if you look at Freshworks, for example, they have customers across EU, US, Australia and other Asia Pacific geographies. And essentially most of the B2B startups on AWS have been on the cloud since day one of their operations. We term them as cloud native businesses, where they are born in the cloud and built off the cloud, and then you have the other set of companies who are now starting to migrate to the cloud. If you look at Freshwoks (a SaaS startup), they have built an artificial intelligence (AI) model for each customer that they work with in order to reduce the response time in terms of customer support…this capability is built on AWS.
With the boom in data, how is your content delivery networks (CDN) service coming into play in India?
A CDN is typically used when you have large amounts of content to be delivered to the end customer without being stored on the public Internet. All of the media and streaming startups use CDNs because they don’t want video content to be stored and played out completely over the Internet there since there will be connectivity issues on end devices. These include Airtel Wynk, ALT Balaji, Saavn and many others. Gaming startups, and even B2B startups catering to clients in the Learning Management Systems (LMS) space have started to utilize CDNs to work around content delivery in patchy networks. AWS currently has around 210 point of presence (PoPs) globally for the CDN operations, and 17 of those PoP locations are now in India, which is the largest CDN presence outside the US market.
Several consumer internet firms in India are embracing audio as an interfacing method, beyond just touch, and it is also being touted as an apt interface for the next set of incoming users in India. Is AWS prepared for this?
Speech recognition is basically three elements: first is the automatic speech recognition (ASR) part of it, then there is the natural language processing (NLP), and the third one is the text-to-speech conversion. What we have done at AWS is that we have taken all those three elements out and made it into three independent services. And on top of this we also have translate service which can then convert text or speech between languages, and we also have comprehend service which can understand emotions and sentiments of the speaker. For India, we have started to add vernacular speech and text recognition in Hindi, Tamil, Telugu, Bengali, and Urdu with more languages in the pipeline.
How are you trying to fix the shortage of skilled data engineers and scientists in India? Are we going to see more data scientist talent moving to freelance basis?
We are doing several things to solve the gap between availability of data scientist talent. So for use cases like text to speech, ASR, NLP, sentiment analysis, we have pre-trained models available. So you really don’t need a data scientist to build the model because the data model is already available from AWS. For example we have made available pre trained models for forecasting, for recommendation engines. The second way is that we are also adding data science talent to our teams. What we essentially do is provide data scientists to our customers on-demand. These are specialists who will work with clients closely, and we are already doing this in India with a bunch of customers.
Can open sourced data sets fix the problem of availability of quality data for building Artificial Intelligence/Machine Learning models?
There are three challenges to performing successful data analytics: first the availability of rich data in big volumes, the second one is the availability of computing power to be able to build that model, and third one is availability of talent pool of data scientists to this job. Recently, we launched the AWS Data Exchange, which essentially makes available data sets to clients across any segments including life sciences, healthcare, telemetry, and public services, etc. Earlier what we did is that we acquired 4 or 5 of these data sets and we made them available, so our clients can use these datasets within AWS and build a new AI Model for them. Now it’s more of a marketplace, where people who have the datasets make it available, to those in need directly on AWS.
The reporter was in Las Vegas at the invitation of Amazon