The world is demanding machine learning more than ever before. Some want to use the technology to analyse big data and discover trends to capitalise on ahead of their competitors. Others want to reduce supply chain wastage. However, due to massive demand, machine learning experts are expensive to hire. Sankha Muthu Poruthotage intends to address this problem and democratise machine learning. “There is a massive demand for machine learning, and we are trying to break the entry barrier with ready-to-use, inexpensive products that can be deployed over the internet,” Poruthotage says.
“There is a massive demand for machine learning, and we are trying to break the entry barrier with ready-to-use, inexpensive products that can be deployed over the internet.”
Firms which subscribe to Linear Squared products do not have to hire teams of data scientists or have a digital consultancy firm on retainer. Most data scientists and machine learning engineers are building customised products for clients. At Linear Squared, Poruthotage focuses on common problems a particular industry has. If the firm had taken a traditional approach, it would have had to employ hundreds of experts, but it manages with 10 data scientists and 17 engineers.
Linear Squared currently has three products, targeting the apparel, retail, and fast-moving consumer goods industries. In apparel, machine learning will plan production, while in FMCG, it forecasts and manages supply chains In retail, Linear Squared addresses the never-ending struggle of targeting the right consumer with the relevant marketing tool. The product breaks down a client list to the granular level. Instead of a handful of segments, a firm would have hundreds of thousands of sub-segments with common threads among these groups identified to customise marketing campaigns.
While a manually-designed product would deliver 60-70% accuracy, Linear Squared punches above the 90th percentile, Poruthotage claims. However, the firm’s products are not 100% flexible due to their mass-market nature. They are built to run on at least 80% of the data which is typical for an industry, but to make full use of Linear Squared products, the remaining 20% of data, usually unique to each client, has to be formatted for use.
For its innovative product, Linear Squared was selected for the Microsoft ScaleUp program in 2019, which will help Linear Squared go global through distribution on the cloud. Linear Squared was built around a consultancy firm Poruthotage founded in 2015. It introduced machine learning products a year ago. Of the 15 clients who have tried the products so far, 13 have stuck on. Linear Squared is generating $100,000 in quarterly revenue, which Poruthotage is hoping to multiply in the years to come.