Dr. Sanjay Basu is a distinguished technology leader with over 30 years of experience in the industry. He currently leads the AI Infrastructure, Generative AI, and GPU Cloud Engineering teams at Oracle Cloud, driving innovation in cutting-edge artificial intelligence and cloud computing technologies.
Dr. Basu holds a Doctorate in Organizational Behavior and a double master’s degree in Management and Computer Engineering. Demonstrating his commitment to lifelong learning and research, he is currently pursuing a second PhD in Artificial Intelligence.
As a recognized expert in the field, Dr. Basu is a Life Member of prestigious professional organizations, including the Association for Computing Machinery (ACM), the Association for the Advancement of Artificial Intelligence (AAAI), the American Association for the Advancement of Science (AAAS), and the Institute of Electrical and Electronics Engineers (IEEE). He is also a Fellow of the Institution of Electronics and Telecommunication Engineers (IETE).
Beyond his leadership roles, Dr. Basu is a prolific writer, having authored five technical books and numerous insightful blog articles. His contributions to technological advancements are further underscored by his six U.S. patents. His thought leadership and expertise in AI, cloud computing, and technology infrastructure are widely recognized in the academic and professional communities.
Optimizing Efficiency and Collaboration in AI Agentic Systems - The Technical Advancements, Team-Based Potential, and Real-World Impact of Small AI Models
Optimizing Efficiency and Collaboration in AI Agentic Systems - The Technical Advancements, Team-Based Potential, and Real-World Impact of Small AI Models
Small language models have emerged as powerful alternatives to their large-scale counterparts, offering advanced capabilities within significantly smaller, more efficient architectures. This paper provides a technical examination of key advancements (e.g. knowledge distillation, pruning) that empower these compact models. Such innovations have dramatically improved small-model performance, enabling them to achieve accuracy and functionality once reserved for much larger systems . For example, DistilBERT retains 97% of its larger predecessor’s language understanding capabilities while being 40% smaller and 60% faster , and some recent small models have even surpassed larger models on specific benchmarks . A major advantage of small models lies in their efficiency and adaptability. They require far less computational power, which significantly reduces energy usage and infrastructure costs . This lean footprint allows deployment on edge devices and mobile platforms without sacrificing responsiveness, expanding access to advanced AI capabilities even in resource-constrained settings . Small models are highly adaptable—they can be quickly fine-tuned for specific tasks or domains, making them versatile tools across a broad range of real-world applications. From intelligent chatbots and on-device assistants to domain-specific analytics in healthcare and IoT, these lightweight models are being applied to deliver fast, context-aware intelligence at scale . Equally important, we explore the collaborative potential of small models operating in team-based AI environments and within hybrid systems. Ensemble techniques demonstrate that even a few small models can collectively achieve accuracy rivaling a single state-of-the-art large model, while maintaining high efficiency . In hybrid configurations, small models can work alongside large models to deliver the best of both worlds—handling specialized tasks locally and invoking powerful large-scale models only when needed . This synergy not only boosts overall system performance but also fosters innovation by combining the strengths of diverse models. Such teamwork and integration strategies are paving the way for new AI solutions that are both robust and resource-efficient, ultimately broadening the impact of AI across industries .