DevConf.IN 2025

Kedar Basutkar

Hi, I’m Kedar! I am an equal parts artist, engineer and builder, currently working as a frontend engineer at Lucidity.

I’m an active hackathon participant and have created apps that solve challenges, from improving internet well-being to enhancing project planning and learning experiences.

As a speaker, I’ve delivered talks on topics like building high-performance web apps and exploring concepts like resumability.

In my free time, I enjoy visiting the theater and appreciating a good drama or comedy.


Company or affiliation

Lucidity

Job title

Frontend Engineer


Session

02-28
16:40
35min
Natural Language Meets Video: Querying Videos Like a Database
Kedar Basutkar

This session introduces a new approach to video search, enabling users to query video content using natural language, much like querying a database. It improves the search process by eliminating manual scrubbing and providing precise results quickly.

The session will cover:

  1. Problem Overview: Challenges with traditional video search and the need for a smarter solution.

  2. Live Demonstration: A real-time showcase of querying videos using natural language and retrieving relevant timestamps and clips.

  3. Technical Insights:
    - How vision models generate descriptive captions for video frames.
    - The use of embeddings to represent both video content and user queries in a shared vector space.
    - Storing and querying this data quickly using a vector database.

  4. Applications: Real-world use cases, including surveillance footage analysis, media curation, and content-aware video search platforms.

This talk is ideal for developers, AI practitioners, and tech enthusiasts interested in understanding the intersection of vision models, natural language processing, and media.

AI, Data Science, and Emerging Tech
Raigad Room (Chanakya Building / School of Business)