Traditional recommendation systems were not enough to address the nuance of how viewers describe what they want to watch. JioStar needed fast, reliable, cost-conscious discovery for entertainment, live-sports moments, and detailed cricket statistics at very large audience scale and across multiple languages.
Reported impact
JioStar built and launched a conversational discovery experience for videos and cricket information across more than 15 languages in about eight weeks, giving viewers a voice-and-visual way to find entertainment, sports moments, and statistics.
JioStar built a conversational video and cricket-stat discovery assistant into JioHotstar using the OpenAI API. Viewers can speak or type requests about movies, shows, live-match moments, or cricket statistics, and receive results through a visual, conversational experience. The team framed production design as a balance of accuracy, latency, and cost, and put answer-scope guardrails in place. Its evaluation approach used a seed dataset expanded with model-generated variations and a maker-checker pattern: one model helped determine an answer while another LLM acted as validator, with domain experts shaping sports-statistics tests. Codex helped generate evaluation cases and update the system. The source does not describe the content-data ingestion, storage, or underlying serving infrastructure beyond the OpenAI API and JioHotstar interface.
Viewers speak or type discovery requests in JioHotstar; the conversational assistant built using the OpenAI API returns video, live-sports, or cricket-stat results in a visual conversational experience.
Codex helped the team accelerate work such as generating eval cases and updating the system
More than 60% of our users use voice rather than text
Using the OpenAI API, JioStar built a conversational video and cricket-stat discovery assistant into JioHotstar.
We worked on bringing entertainment, sports moments and detailed cricket statistics into one experience, across multiple languages, in a very short period.
one model helped determine how to answer a question, while another acted as a validator—an LLM as judge
Review how this reference was scored and trace each normalized claim back to its documented source.
Scoring methodology
Customer Reference Value measures usefulness as a customer reference. Evidence Confidence measures the quality and traceability of available evidence, not independent verification.
Traceability
Field coverage shows which facts each source supports. Dates are shown only when documented by the source.
Related
OpenAI Sample Oct 8, 2026 Axios AI is already a huge part of how Axios Local works. At the core, what we’re trying to do is prove that you can run a sustainable, profitable local news model that delivers high-quality journalism to every community in… Axiomizer · custom GPTs OpenAI Sample Oct 8, 2026 CyberAgent CyberAgent needed a secure, governed way to spread AI across advertising, media, gaming, and development teams so employees could work faster and make better decisions without risking uncontrolled data use or fragmented… 93% monthly active usage of ChatGPT Enterprise ChatGPT Enterprise · Codex OpenAI Sample Sep 9, 2026 ATV Big Air Tour ATV Big Air Tour needed a way to handle show logistics, marketing, communications, and on-site problem solving with a very small team because administrative work was consuming time the founders needed for growth and… Most show days they were working until 3am just to keep up ChatGPT · ChatGPT voice mode 