Meesho declared at the Nvidia Summit in Mumbai that it is making its cutting-edge machine learning (ML) platform open-source in an effort to empower data scientists and developers everywhere. The move would improve AI accessibility and creativity, according to Debdoot Mukherjee, head of AI at Meesho, who described the open-source release in detail, including the company’s feature store and model orchestrator.
He emphasised Meesho’s sophisticated machine learning architecture in his talk on cost-effective, high-performance model serving. It is designed to serve millions of requests per second, with 99 per cent of them handled in less than 10 milliseconds. With ambitions to open-source this technology, this effective Feature Store is set up to connect raw data with ML models and optimise data pipelines.
Meesho’s ML integration goes beyond enhancing the seller experience by using generative AI to help with listings and translate content between Indian languages. Additionally, Meesho has improved call centre operations and raised customer satisfaction by 20% with the help of Microsoft’s Azure Open AI.
With its open-source machine learning platform currently tailored for large e-commerce operations, Meesho is well-positioned to give developers the resources they need to create scalable machine learning systems for the sector.
