AWS AI Conclave shows why you don’t need a PhD but got to be a builder at heart to innovate with AI
Amazon and industry experts showcase how brands and Amazon architects are building, training and deploying sophisticated models and unlocking intelligence, powered by AWS technologies
“You don’t need a PhD; you need to be a builder at heart,” said, Denis Batalov, Worldwide Technical Leader, ML & AI, AWS in his keynote address on day 2 of the Amazon AI Conclave in Bengaluru.
The Artificial Intelligence (AI) expert with over two decades of experience took the audience through how AWS developed comprehensive ML capabilities and solutions over the years and how it has reduced the barriers for innovation.
“AWS ML Stack is probably the broadest and most complete set of ML capabilities. Today, we have tens and thousands of customers who have chosen AWS for ML workloads. But, the journey has only begun.”
Denis further deep-dived into AWS’ AI services and highlighted how Amazon’s foundational work that started many years ago led to innovations like Sagemaker.
“Today, Amazon makes it easier to build, scale and apply AI services.”
An obsession with innovation
If day 1 of Amazon AI Conclave put the spotlight on the developments in AI led by AWS and recognised startups doing some revolutionary work in the area of AI, day 2 of the conclave focused on some of the hottest stories of both AWS partners and architects.
Designed as the technical edition, Day 2 provided a platform for Data Scientists, ML developers, data engineers and architects to hear from thought leaders across fields and learn how they were applying some of the best practices in AI and ML.
Denis was followed by Akanksha Balani, Region lead - Intel® Software, who spoke about how Intel’s partnership with AWS, is making it easy for the tech ecosystem to access Intel’s computer capabilities. She also highlighted how the recently unveiled Intel’s One API makes it easier to write AI code once and run anywhere.
Up next, Sohan Maheshwar, Evangelist, Alexa Skills, deep-dived into one technology - voice computing and Alexa. In his energy-packed informative session, Sohan took the audience through some of the innovations in Alexa, its capabilities and tools available for developers to come up with such innovations.
Big opportunity for ML enthusiasts - AWS DeepComposer and AWS JPL Open-Source Rover Challenge
Girish Patil, Solutions Architect (AI/ML), Amazon India, demoed AWS DeepComposer keyboard to create a melody powered by AI and highlighted how the musical keyboard and ML techniques provide a creative way for developers build generative AI models, all without having to write a single line of code.
At the conclave, he asked developers to participate in the AWS JPL Open-Source Rover Challenge – a virtual hackathon which runs through February 21, 2020. The Challenge is a call for tech enthusiasts to use their ML skills to build and train a reinforcement learning (RL) model on AWS to autonomously drive JPL’s Open-Source Rover between given locations in a simulated Mars environment.
Know more about the challenge here.
From logistics to customer engagement: How AWS Partners are solving key business and social issues to drive impact
The conference also featured 15+ breakout sessions across four tracks delivered by Amazon and industry experts that showcased how brands and Amazon architects are to building, training and deploying sophisticated models and unlocking intelligence, powered by AWS. Here are a few highlights from some of the sessions:
Improving customer experience
Madhu Gopinathan, Senior Vice President, Data Science, MakeMyTrip, shared how AI-powered chatbot Myra is helping improve customer experience by providing fast and contextual response to not just regular customer queries but also complex ones. “The bot has been built to understand nuances such as WhatsApp lingo, Bi-lingual messaging and also deal with intent ambiguity. Today, of the 12K customer chats that the platform handles, 80 percent of that is handled by the bot.” MakeMyTrip uses standalone Sagemaker Notebook for each Data Scientist instead of project-specific beefy machines. “This helps to avoid resource contention. It also helps to scale up or down compute and memory for tasks such as decouple analysis versus training versus inference. Sagemaker also helps to bring down run time costs by putting the different models in the same container.”
Kabir Rustogi, Head - Data Science, Delhivery, showcased how they optimised the supply chain network by letting go of the reliance on the Pincode system and building an AI-led solution. He showcased how their in-house platform, Addfx, can not only fix the address when certain details are missing or incorrect, but also helps to pinpoint the address with great accuracy. Another AI-led product helps to automate dispatch and create a load plan for field executives, which helped reduce dispatch creation time. Using the Amazon SageMaker Single-Shot Multi-Box Detector algorithm helped Delhivery to automate the Proof of delivery verification and reduce the manpower requirement by 60 per cent.
The best in the game
Sandeep Agarwal, CTO, Games24x7, explained how Reinforcement learning and applications at Games24x7 is aiding personalisation and in the process of creating the world’s best rummy player. “Today, ML is bringing inefficiency in use cases such as landing page optimisation, new feature introduction, creatives for push notifications, automating media buying, registration optimisation to even in table assignment to match people with similar skill sets.”
The secret sauce in on-time food delivery
Avinash Ruchandani, Product Manager, Zomato, shared how they are acing the delivery time prediction by predicting time taken for even a single leg of the journey with an ML- based solution, “One of the biggest challenges is to run these models on low latency. But, Sagemaker solves this challenge. In addition, Sagemaker is simple to use and can be handled end-to-end by data scientists themselves.”
Optimising customer engagement
Handling 700K requests a minute and powering eight million conversations a day, Freshworks operates at a massive scale. Swaminathan Padmanabhan, Director, Data Science, Freshworks, said,“There’s a need to build, deploy and manage tens of thousands of models simultaneous to power every AI/ MLinitiative at freshworks. We leverage multiple AWS technologies like Sagemaker, Lambda, Step function, to achieve this.”
Achieving and managing scale: Paytm and Chaayos
"We knew that our app needs to understand every user and their journey intimately. We have four petabytes of consumer data and approximately 13000 data touch points on every user", said Ankit Sinha, VP, Paytm sharing AI's power to reach right user segments during his talk.
Highlighting the importance of technology in scaling, Mohit Malik, CTO, Chaayos, said, “The whole-sole idea of marrying tea and technology was to build every tech product inhouse which helps in scaling. If you have to be ahead in the game, you have to invest in tech.”
The event also featured a number of hardcore technical tracks focused on AWS technologies that help to build, deploy and scale AI solutions and niche AI topics such as speeding up Deep Learning training and inferencing, cutting cost & improving performance and automating code reviews and application performance recommendations, among a host of others.
The end of Day 2 of the Amazon AI conclave essentially gave an opportunity to developers, data scientists and tech professionals to understand how growing businesses in India are building smart, customer-centric, scalable solutions and how AWS supports them in this endeavour.