Job Description
Overview
Job Description:
This role involves spearheading the development of an intelligent video analytics and search system. We're building a solution that takes vast amounts of video footage, whether streaming live or from existing archives, and meticulously processes it. This means segmenting videos into individual frames and then applying advanced AI models: object detection and classification to identify distinct items within each frame, and cutting-edge Video Language Models (VLMs) to provide rich, natural language descriptions of these objects and their attributes (for example, instead of just man and umbrella, a VLM might describe a man holding a black umbrella). All this generated textual information is then meticulously organized and stored in a specialized vector database, optimized for incredibly fast and intelligent retrieval. When a user needs to find something specific, like find the man with a black umbrella, they simply input their query in natural language. Our system then leverages this vector database to efficiently pinpoint and present the exact, relevant video frames, effectively transforming immense video datasets into actionable, easily searchable archives.
Location: Bangalore/Chennai
Responsibilities
Job Description:
This role involves spearheading the development of an intelligent video analytics and search system. We're building a solution that takes vast amounts of video footage, whether streaming live or from existing archives, and meticulously processes it. This means segmenting videos into individual frames and then applying advanced AI models: object detection and classification to identify distinct items within each frame, and cutting-edge Video Language Models (VLMs) to provide rich, natural language descriptions of these objects and their attributes (for example, instead of just man and umbrella, a VLM might describe a man holding a black umbrella). All this generated textual information is then meticulously organized and stored in a specialized vector database, optimized for incredibly fast and intelligent retrieval. When a user needs to find something specific, like find the man with a black umbrella, they simply input their query in natural language. Our system then leverages this vector database to efficiently pinpoint and present the exact, relevant video frames, effectively transforming immense video datasets into actionable, easily searchable archives.
Requirements
Skills:
computer vision - batch and streaming video analytics, multi-object detection and classifications, VLMs
NLP - LLM and building RAG S/M.
Expertise in Python
Knowledge of AI & ML
Cloud & Infrastructure - Docker, Kubernetes, kafka
APIs
