German I. Parisi
Program Director, Product Management, IBM watsonx, New York (NY)
Program Director, Artificial Intelligence, IBM Data & AI, San Jose (CA)
Co-founder and former Board Member, ContinualAI
Program Director, Product Management, IBM watsonx, New York (NY)
Program Director, Artificial Intelligence, IBM Data & AI, San Jose (CA)
Co-founder and former Board Member, ContinualAI
I’m passionate about developing innovative and impactful systems that are interactive by design and delightful by experience.
I currently work on generative AI and agentic frameworks at IBM watsonx, where I design and develop AI/ML-powered algorithms that learn from millions of real-world human-agent interaction examples.
I believe in the transformational power of quantum computing and its incredible potential to disrupt and advance AI.
Some background:
In 2019, our Mountain View-based start-up Apprente Inc. was acquired by McDonald's Corporation with the mission to advance state-of-the-art ML, AI, and related technology solutions that address real-world, data-driven needs in the McDonald’s Restaurant environment and completely transform the customer experience (Press coverage: TechCrunch - WIRED - Business Insider - Yahoo - Forbes).
My 2020 interview on Fortune about the future of QSR and drive-thru automation, and on McD Tech Labs' mission and projects: WIRED, CNN.
I'm the co-founder and former Board Member of ContinualAI, the largest non-profit research organization on continual learning for artificial intelligence with over 2000 active users, online seminars, sponsored workshops, and the community behind Avalanche, an open-source CL library based on PyTorch.
In 2021, my co-authors and I received the Best Article Award from the International Neural Network Society for our paper on continual learning.
In 2021, our R&D division at McDonald's Corporation, McD Tech Labs, was acquired by IBM, and in 2022 our voice-activated solution became part of the watsonx portfolio.
Churamani, Dimlioglu, Parisi, Gunes. (2023)
Continual Facial Expression Recognition: A Benchmark.arXiv preprint arXiv:2305.06448