My Work on Brand Safety and Suitability in Social Media at Integral Ad Science
Welcome to My Blog on Brand Safety and Suitability on Social Media
This blog highlights my work as featured in IAS newsletters and recognized across the AdTech industry. In this blog, I will share insights and my involvement as a member of the video classification team, machine learning solutions team, and multimedia enablement team at Integral Ad Science (IAS).
Advancing Brand Safety and Suitability on TikTok with Machine Learning at IAS
INTRODUCING IAS PRE-BID BRAND SAFETY ON TIKTOK
My work at Integral Ad Science (IAS) leverages advanced machine learning technologies to ensure brand safety and suitability on TikTok across major markets including the US, Canada, UK, Germany, France, and Australia. Utilizing proprietary frame-by-frame video, audio, and text classification technology, IAS provides unparalleled transparency and granular control for advertisers. This innovation allows brands to confidently place their ads, maximizing their impact.
I am a key member of the video classification team at IAS. My primary role involves training and deploying deep neural network models that perform video multimodal signal aggregation and classification in accordance with the Global Alliance for Responsible Media (GARM) rules and guidelines; more on the GARM guidelines here: GARM guidelines. Through our work, IAS secured a partnership with TikTok to develop third-party brand safety and suitability solutions, ensuring that advertisements are placed near high-quality, safe, and suitable content.
My Role in Expanding Coverage of Brand Safety and Suitability on TikTok
POST-BID BRAND SAFETY & SUITABILITY ON TIKTOK
In addition to the pre-bid brand safety solutions, my work has expanded to include post-bid solutions, which handles significantly higher daily volumes. My focus remains on multimodal signal aggregation and expanding text classification to cover many international languages. This comprehensive approach provides end-to-end transparency and protection for entire TikTok campaigns, aligning with GARM Brand Safety and Suitability categories.
Comprehensive Solutions for Identifying Misinformation Narratives on Social Media
Additionally, My role at the Multimedia Enablement team at IAS includes developing solutions to tackle harmful misinformation on social media. My work involves ideation, team formation, hands-on implementation, and evaluating large language models (LLMs) with retrieval-augmented generation (RAG) architectures and tool-using agents. These technologies accurately identify misinformation topics and provide interpretable reasoning around facts and fake news. The impact of this work allows IAS to confidently detect misinformation, a crucial capability that clients are seeking in the election year 2024.
Indutry-First Deepfake Measurement Capabilities
IAS ANNOUNCES INDUSTRY-FIRST DEEPFAKE MEASUREMENT CAPABILITIES
Furthermore, my project at the Multimedia Enablement team at IAS include developing deepfake detection capabilities, using AI-powered technology My direct involvement includes ideation, creating proof-of-concept, moving to MVP version, productionizing, and post-launch evaluation. Our deepfake detection technology uses AI-driven multimedia technology with frame-by-frame analysis to detect visual manipulations, such as faceswaps in videos and images. This innovation enhances our misinformation categorization and brand safety products, providing advertisers with the most reliable and accurate data to safeguard their brands.
P.S. While this blog showcases my successful work that has become products and garnered external recognition, it is a result of the collaborative efforts of the talented data scientists and dedicated engineers of our team, and reflects the strong, curiosity-driven innovation culture at IAS. There are numerous experimental proof-of-concept projects and internal-facing efforts that are equally important in my opinion. These endeavors are fundamental to innovation and continuous improvement.
Thank you for visiting, and I hope you find my work as fascinating and impactful as I do. If you have a question, please reach out!