Milind Naphade

Milind Naphade is an AI researcher and technology leader who joined Capital One in May 2023 to build its AI Foundations organization. By June 2024, he was SVP, Technology, and head of AI Foundations, following leadership roles at IBM Research, Cisco and NVIDIA, where he led technology strategy for the Metropolis platform. His work spans interpreting multimedia, connecting technical-support cases and applying AI to business problems.[2][4]

Naphade described his move into financial services as a response to ChatGPT’s November 2022 release. He expected innovation to shift from underlying platforms toward solving business problems, and believed that required being close to the relevant data and context. Financial services, he said, was sufficiently closed to outside access that meaningful innovation required working inside the industry. In his June 2024 interview, he put the requirement plainly: “the only way for me to be part of that was if I was close to the business data and context needed to solve those problems.”[2]

During his first year, Naphade built AI Foundations, established partnerships with Capital One’s business and technology groups, and contributed to its AI strategy. Alongside the wider technology organization, he helped launch an Applied Research job family and a related Ph.D. internship program. He described research priorities ranging from anomaly detection and behavior models to large language models and graph networks, with understanding the business informing experiments and prioritized use cases.[2]

In June 2024, Naphade endorsed Yann LeCun’s vision of autonomous machine intelligence built around learned internal models of how the world works. He believed reasoning based on such world models would be essential to the next major shift in AI, enabling machines to learn faster, plan complex tasks and adapt to unfamiliar situations. He saw particular relevance to financial decision-making.[2]

He also expressed interest in smaller language models, including Mistral/Mixtral and the Llama 3 series, for potentially more efficient and scalable financial applications. He viewed multimodal LLMs that interpret text, audio and images and exchange outputs across media as a way to enrich customer interactions. These were his June 2024 assessments of emerging capabilities.[2]

Naphade credited NVIDIA’s “speed of light” ethos—setting benchmarks aside to pursue the best possible performance—and its intellectual candor about successes and failures as influences on his career. Reflecting on Metropolis, which used sensors to monitor and optimize physical environments, he argued that “impactful AI platforms must be built with humans as their focus.”[2]

Before his move to Capital One, Naphade’s work included multimedia research, urban mobility analytics and customer-support systems. Those earlier projects provide concrete examples of the AI platforms and applications he brought to his financial-services role.[2][4]

Naphade earned a Bachelor of Engineering in instrumentation and control from the College of Engineering Pune, followed by an M.S. and Ph.D. in electrical and computer engineering from the University of Illinois at Urbana-Champaign.[4]

With Igor Kozintsev and Thomas S. Huang, he coauthored Probabilistic Semantic Video Indexing, published in the NIPS 2000 proceedings. Their framework mapped low-level media features to semantic labels through probabilistic multimedia objects called multijects. Networks of those objects, called multinets, represented relationships between concepts, including their co-occurrence and temporal dependencies within video shots. Using factor graphs and sum-product inference, the researchers sought to correct isolated concept-detection errors by enforcing contextual constraints.[5]

During 14 years at IBM Research, Naphade led AI initiatives in multimedia analytics and smarter-planet applications. He recalled that IBM’s Insights in Motion mobility platform began with a chance conversation at the White House: an unnamed assistant city manager asked whether mobile data could help revitalize her city’s bus system. The team developed a platform that, in his account, analyzed movement using billions of anonymized telecommunications records and supported transit optimization and location-based marketing. He credited it with award-winning transit implementations in Dubuque, Iowa, and Istanbul, Turkey.[4][2]

At Cisco, Naphade directed the Cognitive Systems Organization and built a team focused on machine learning, natural language processing and AI. He recalled that the team used embeddings and a homegrown vector database to discover connections between technical-support cases, helping the technical services team resolve customer issues.[4][2]

At NVIDIA, he served as chief technology officer of AI Cities and led Metropolis technology strategy. He also coauthored The 7th AI City Challenge, submitted in April 2023, with Shuo Wang, David C. Anastasiu, Zheng Tang and other collaborators. The challenge connected computer-vision research with retail and traffic applications. Its new multi-camera people-tracking track used both real and realistic synthetic data; it also separated a contest leaderboard that excluded external private data from a general leaderboard for all submitted results.[4][6]