JioStar’s chief architect on AI discovery at 500M-user scale

JioStar, which operates entertainment and live-sports platforms serving over 500 million consumers, built a conversational AI discovery assistant into JioHotstar using the OpenAI API. The assistant supports natural voice and text queries across more than 15 languages, allowing viewers to find movies, shows, cricket highlights, match moments, and detailed cricket statistics through a visual, conversational experience. At OpenAI's Executive Summit in Singapore, Vijay Seshadri, JioStar's Chief Architect, spoke with OpenAI's Sidharth Sharma about developing the experience in about eight weeks for a major cricket tournament, serving both entertainment viewers and cricket fans. Seshadri emphasized a three-way tradeoff among accuracy, latency, and cost; noted that over 60% of users prefer voice; and highlighted the importance of domain experts in evaluation, guardrails that define what the system should not answer, a maker-checker validation pattern using one model to answer and another to judge, and production monitoring of model efficacy. He also advised executives to focus on outcomes rather than activity and to recognize the human judgment and skilled people required to move AI from demo to production.

(Source：OpenAI)