AutoML is a popular application to automate machine learning model development, allowing data scientists, analysts and developers to build high-accuracy, high efficiency ML models while sustaining the quality of the models. With no humans required to intercede on new queries, AutoML leads to the creation of actually ‘smart’ systems. This branch of AI is hugely popular where reams of data is involved, so it’s an ideal fit in sectors like banking, insurance and supply chain. The efficacy of ML systems depends on data integrity, eventually leading to innovative solutions and products being developed on real customer data.
Data-rich sectors do have a lot to benefit from AutoML, but the benefits of this technology are not restricted solely to such sectors. This was the opportunity Rahul Vishwakarma and his college mate Kailash Ahirwar chanced upon in their college days. From helping companies with their AI/ML integration strategies to contributing to open-source projects involving deep learning frameworks, Vishwakarma and Ahirwar wanted to enable businesses and non-developers to adopt machine learning. “I am convinced that AutoML technology is extremely powerful for AI start-ups and companies involved in analytics,” said Vishwakarma, cofounder and CEO, Mate Labs.
They launched Mate Labs, India’s first horizontal AI platform, in 2016. In two years, the duo launched their flagship product Mateverse, with the sole purpose of simplifying the process of building and training ML models – so much so that it would be simple for someone who had never written a line of code. “We took a B2C approach, believing that launching with a no-code platform would be a good way to get started, and by 2018, we had over 30,000 users,” said Vishwakarma.
While the B2C approach validated the market approach, conversations with industry leaders and larger corporations made Ahirwar and Vishwakarma realise something. “Even if we make the platform extremely simple to use and learn ML, people will not have time to build anything on it on their own.”
By the end of 2018, the duo made the decision to automate an entire use case rather than the process of building ML models. They experimented and piloted with five large companies from banking, insurance, supply chain, telecommunications, and even the government that were actively looking to use these advancements. Following a few experiments, majority of the impact they could tangibly expect was in the supply chain sector, which was still using traditional methods, was data-rich and ready to be disrupted by innovation.
The duo worked on their first pilot project with a leading pharmaceutical company and introduced demand forecasting, which lies at the heart of any manufacturing or consumer goods company. “Forecasting is one of the most difficult of the four broad challenges in machine learning/deep learning. When attempting to predict something that has not yet occurred, there are hundreds of thousands of factors that can sway your predictions, which is where we saw it as a good challenge and decided to solve it. Within the first pilot, we were able to provide a 30% increase in accuracy using our platform,” explained Vishwakarma. Mate Labs claims it has built the world’s fastest AutoML technology, which bridges the gap between supply and demand by allowing supply chains to reduce response time from 3 to 4 months to near real-time. Deployment time is just 15 minutes versus the standard 9 to 10 months. For instance, a leading CPG company required an offline demand forecasting for all SKUs with a guaranteed accuracy jump. Before using MateVerse, the client’s accuracy rates were between 45-50%. After using Mateverse, these rates climbed to 70% with projected savings of $1 million. While Mate Labs has automated the existing demand planning process, they are now working to automate inventory management, vendor management, distribution, and logistics.
Unlike SaaS companies, whose proprietary CRM tools can be customised to integrate with existing legacy systems, AI companies, sell decision making or intelligence. With a lot more at stake, there is a huge opportunity for companies to use AutoML technology to significantly speed up or optimise their delivery pipeline, explains Vishwakarma.
Mate Labs currently helps fortune 500 companies with the fastest and most accurate demand forecasts. Customers include L’oreal, Marico and Luminous among others. In 2019, Google recognised Mate Labs’ AutoML technology as the fastest in the world, allowing them to create a more accurate and near real-time forecasting system. Vishwakarma adds that they will not raise invoices if they cannot maintain the accuracies promised to clients – a first in the industry.
Currently, the startup is working with global Fortune 500 companies to develop next-generation supply chains that can be fully automated. Eventually, a single supply chain control tower should manage the entire supply chain. They’re already working with a leading global beauty and cosmetics brand to reduce waste in their supply chain by developing a first-of-its-kind solution based on SLOB (Slow Moving and Obsolete Inventory). Additionally, they have also created a complete automated demand planning system that opens up a slew of opportunities for inventory, distribution, and manufacturing.
Source: indiaai.gov.in
