Optimizing content analysis with AI and data analytics for a leading media technology company

The client’s content analysis team started receiving an increasing number of requests for hundreds of thousands of media assets (images, videos, audio, and other files that have to adhere to specific requirements) from customers that were training various AI models. The client wanted to streamline the process of handling such requests, removing operational overhead from their time and making sure that all customers were served without delay.

Our data engineering team, together with the client, analyzed their extensive library of over 1.5B media assets to identify the images, videos, etc. that matched the specific needs of each customer. We used ad-hoc Snowflake-based SQL queries to extract specific metadata from different assets, parse it, and prepare the necessary details, such as keywords or descriptions, for the client’s team.

Vector similarity search was utilized to find relevant content by capturing essential data features in a structured and measurable format.

Furthermore, our engineers developed a PoC for content analysis automation. We used Python libraries to recognize images, and the Databricks analytics platform, powered by machine learning, for quick and effective data processing.

• Enhanced service quality by enabling the client’s team to quickly address requests for large amounts of images, videos, audio, etc. from a library of over 1.5B assets;
• Accelerated asset search by 100 times by utilizing the vector similarity search.
• Streamlined content analysis with a new tool powered by AI and machine learning;