Bhavesh Upadhyaya - “Agentic Metadata Enrichment: Why Human-in-the-Loop Is the Only Path to Trust”
Автор: MonteVIDEO Tech
Загружено: 2026-02-10
Просмотров: 27
Описание:
“Agentic Metadata Enrichment: Why Human-in-the-Loop Is the Only Path to Trust” Bhavesh Upadhyaya,
This talk was recorded as part of Montevideo Tech Day, during Summer Camp 2026.
In this session, Bhavesh Upadhyaya takes a concrete, code-level look at using AI agents for metadata enrichment in real media workflows. He examines why hallucinations don’t simply disappear with better prompts or larger models, and how assuming “AI magic” often leads to failures rooted in incomplete instructions and flawed reasoning. The talk makes the case for human-in-the-loop feedback as a core reinforcement mechanism, focusing on measurable confidence: capturing corrections, reinforcing agent behavior, and iterating until metadata enrichment can be trusted at scale.
Montevideo Tech Day brings together technical voices from across the video industry to share perspectives and experiences around the end-to-end video workflow, through short tech talks, lightning talks, and roundtable conversations.
📍 Summer Camp 2026
Montevideo, Uruguay
January 2026
Повторяем попытку...
Доступные форматы для скачивания:
Скачать видео
-
Информация по загрузке:
