For NoBroker, Google Cloud enabled the development of an array of solutions on its application in a short time during the peak pandemic period and automate many processes.
One of the solutions developed was a grocery delivery system on its NoBrokerhood app, which manages services in large apartment complexes. The feature enabled residents to place grocery orders, which would be aggregated and sent to BigBasket. The online grocery chain would then send one of its delivery agents to cater to the apartment.
“We could do this with ease on Google Cloud,” says Akhil Gupta, co-founder and CTO of NoBroker. The company also quickly developed a walkie-talkie feature on the app to enable security guards to communicate with one another since social distancing had become mandatory.
These features were built using Firebase, a Google platform for creating mobile and web applications. “If I had to build my own application for this, it would have taken a few weeks,” says Gupta.
Google Cloud has also offered the huge advantage of automatic scaling, up and down, of server usage, depending on the traffic. This was possible through Kubernetes. In the past, scaling resources on the legacy system was a manual process for the engineering and the DevOps teams and required almost an hour.
“It was always a challenge when on weekends traffic is high and on weekdays low. And the site would sometimes break. Once we were on Google Cloud, we did not have to worry about the traffic flow as when it was low, as during the pandemic, we knew our machines were already scaling down and saving costs,” Gupta says.
NoBroker also uses the Google Kubernetes Engine for continuous integration and deployment (CI/CD) to speed up development time. CI/CD pipelines run automated workflows such as code checks and make changes if necessary. Gupta says it’s easy for developers to build, test, and deploy applications at scale on Google Kubernetes Engine without worrying about configuration. “We reduced 80% of maintenance support time,” he says.
Google TensorFlow has helped improve the curation of pictures of apartments beside every listing. Owners would often post pictures which had no relevance with the listing and NoBroker had a team of 35 people to scan through all pictures to remove the irrelevant ones. This was cumbersome. The TensorFlow-based solution identifies objects within a picture – such as whether there is a bed, kitchen, sofa set – to identify whether the image is of a property. “We were able to reduce the size of the team scrutinising the pictures,” Gupta says.
NoBroker also brought to market in three weeks another feature using TensorFlow and Google AI/ML, called Touchless Entry, that uses facial recognition to detect residents of a property, including household staff. This aims to protect residents by helping them reduce physical contact with surfaces where the virus might live.
Gupta says at one point they realised their core platform was on one infrastructure, and all the rest on Google Cloud. “And we thought, why not have a single cloud infrastructure that takes care of all our requirements. So now we are fully on Google Cloud,” he says.