Meesho open-sources high-performance ML platform at Nvidia Summit in Mumbai
28 October 2024, Mumbai
At the Nvidia Summit in Mumbai, Meesho announced it is open-sourcing its advanced machine learning (ML) platform, a move aimed at empowering developers and data scientists globally. Debdoot Mukherjee, Meesho’s head of AI, detailed the open-source release, including the company’s feature store and model orchestrator, stating that the initiative will enhance AI accessibility and innovation.
In his presentation on cost-effective, high-performance model serving, Debdoot Mukherjee highlighted Meesho’s advanced ML architecture, engineered to handle large-scale, low-latency requests by processing millions per second, with 99 per cent managed in under 10 milliseconds. This efficient Feature Store is structured to streamline data pipelines and seamlessly connect raw data with ML models, with plans for open-sourcing this technology.
Meesho’s ML integration extends to improving seller experience through generative AI for listing assistance and translations across Indian languages. Further, with support from Microsoft’s Azure Open AI, Meesho has optimized call center operations, achieving a 20 per cent boost in customer satisfaction.
Meesho’s open-source ML platform, already optimized for massive e-commerce operations, is poised to provide developers with tools to build scalable ML systems across the industry.
