The Amazon Web Service wants to drive the public sector partnership in India with effective Artificial Intelligence and Machine Learning tools.
For Amazon Web Service, public sector partnership is an increasingly important one in India. The world's leading vendor in public cloud services already counts important agencies and ministries in India as its customers and is hoping it's Artificial Intelligence (AI) and machine learning (ML) tools can help build more applications to improve public services in India.
AWS is also impaneled with Maharashtra government and hosts mission-critical applications and databases for them, including the Farmer Loan Waiver System, the Maharashtra Real Estate Regulatory Authority and the Direct Benefit Transfer on AWS. But among AWS' array of services, the newest tools are ones with a focus on AI and ML.
According to Sehgal, these hold tremendous potential for India's public sector. "We don't just offer machine learning as a service. It is built into a lot of our existing services with AWS, including analytics. Customers like Common Service Centers (CSC) are using AWS to actually innovate on the ground," Sehgal said. Common Services Centres are the government of India mission providing e-services at the village level.
He gave the example of AWS' recently launched Textract service, which relies on AI tools and could make digitization easier at these CSC. Textract can extract the text from digitally printed forms and even scanned documents, making it more machine-readable. "A lot of forms, which are printed and scanned are sent over to these CSCs. Here the villagers do the digital entry. There is AI which can help in between. Textract takes a form, printed or even a hand-written, and analyses it in real-time. It doesn't just get the data. It also figures out whether there's a table," he explained.
For the government, a service like Textract could make it a lot easier to extract data inside forms, rather than relying on manual entry alone and thus speed up digitizations. "If you really want to do very use-case specific customized machine learning, then you have to have training data, you have to build a model. So we have launched a service called SageMaker Ground Truth, which does that. So it actually takes training data, and does auto labeling, to whatever extent it can to help build the model," he said.
But one of the challenges when it comes to creating AI and ML services specifically for India is the lack of open datasets in the country. According to Sehgal, open-data sets can help the public sector, and even academia when it comes to building specific applications, what he called the â??missing ingredient'.
"You do have data.gov. But if you go and search within data.gov, and you will most likely find data sets, which are a summary of a dashboard. The raw data, which is what is needed for machine learning is missing," he said. Sehgal gave the example of a machine learning ready data set from India that has been put on AWS' Open Data set registry from Janaagraha, which has citizen complaints around the Swacch Bharat mission.
"It's a very image-heavy data set, which can be used for computer vision applications. We are in talks with a number of other public institutions and academia to bring in their data sets, possibly even Ministry of Urban Affairs. They could be certain city level data, which can be desensitized for instance," he said, adding that such data can be easily anonymized in order to create ML applications based upon it.