Surabhi Bhargava | Source | Adobe
Surabhi Bhargava
Surabhi Bhargava is a seasoned Machine Learning Tech Lead at Adobe with extensive experience in building and scaling AI products. She currently leads the development of Generative AI features in Acrobat which incorporates innovative features such as Q&A across multiple documents, Retrieval-Augmented Generation (RAG) systems, large language model (LLM) evaluation, and intent detection for the AI Assistant in Acrobat. Surabhi has a proven track record of launching successful AI-powered products, including Adobe’s Liquid Mode and Extract, both recognized as one of the Top 100 Innovations by Time in 2023 and attracting over 25 million monthly active users.
With a strong background in productionizing machine learning research, Surabhi specializes in Generative AI, Computer Vision, and Natural Language Processing (NLP). Her contributions are also recognized within the research community, with her work amassing over 1000 citations. She also serves as a reviewer for prestigious conferences like ACM MM, NeurIPS, AISTATS, and esteemed journals such as Elsevier.
Beyond her technical background, Surabhi is passionate about mentoring the next generation of AI professionals and actively shares her knowledge through international conferences and by guiding students and early-career professionals.
-
Understanding AI: Key Fundamentals for Developers
Surabhi stresses the importance of understanding "model dependency on data" and evaluation techniques. Developers must recognize AI's limitations and biases. Without proper knowledge, organizations risk unmanaged biases and misaligned metrics. Surabhi advises leveraging resources, experimenting with open-source models, and engaging with AI communities to enhance understanding and implementation. -
GenAI Revolutionizes Programming: Efficiency and Caution Key for Developers
Surabhi highlights increased efficiency with GenAI in tasks like prototyping and debugging, while cautioning against over-reliance. She advises developers to embrace AI but remain hands-on, ensuring AI-generated code meets project needs. Surabhi emphasizes, "AI is a tool, not a replacement for expertise," stressing the balance between automation and expertise for innovation. -
AI Watermarking: Distinguishing Authenticity in Digital Content
AI watermarking emerges as a solution to verify content authenticity and protect intellectual property. Challenges include model evolution and enforcement, with alternatives like AI detection methods and governance on content distribution.