580 California St., Suite 400
San Francisco, CA, 94104
Key research themes
1. How can modular AI pipelines facilitate end-to-end integration of AI on resource-constrained embedded devices (Edge AI)?
This research theme investigates frameworks and methodologies for integrating data ingestion, model training, deployment optimization, and IoT integration in an end-to-end manner to enable efficient AI deployment on embedded and edge computing platforms. It addresses the technical and organizational challenges small and medium enterprises face when adopting AI on constrained hardware, focusing on modular pipeline design and deployment optimization strategies to improve accuracy and operational efficiency without necessitating extensive AI expertise.
Key finding: This paper introduces a modular AI pipeline developed within the European Bonseyes project that integrates data ingestion, model training, deployment optimization, and IoT hub integration into a cohesive framework targeting... Read more
2. What is the evolving role of human actors in the AI loop within sociotechnical production environments to enhance AI capabilities?
This theme explores the interaction between human workers and AI systems in industrial production setups, emphasizing the collective activities emerging from human-AI collaboration. It investigates how human cognitive capabilities, situational awareness, and active involvement can augment AI effectiveness beyond simple augmentation, influencing design paradigms that integrate human contributions into AI decision-making loops, thereby improving operational performance, adaptability, and work design considerations.
Key finding: This study presents a conceptual model and empirical case study demonstrating that involving humans actively within AI systems on production lines—specifically in vision-based inspection tasks—can significantly enhance AI... Read more
Key finding: This paper synthesizes management challenges and opportunities related to integrating AI into organizational contexts, emphasizing human-AI interaction governance. It identifies the necessity for managers to balance AI’s... Read more
Key finding: The paper documents the historical divergence and current convergence of AI and robotics research, emphasizing the renewed focus on embodied intelligent systems that integrate perception, reasoning, and actuation. It argues... Read more
3. How can language-agnostic sidecar microservices enable AI integration into legacy enterprise systems without disrupting existing infrastructure?
This research area investigates architectural patterns and protocols to integrate AI capabilities, especially Large Language Models (LLMs), into established enterprise platforms (such as .NET Core) without rewriting or compromising existing applications. It focuses on microservice-based sidecar architectures that locally host AI inference engines and communicate via standardized, language-agnostic protocols, enabling semantic enrichment, intent classification, and data transformation within regulated, secure environments where cloud dependencies are restricted.
Key finding: This paper outlines a novel system architecture employing a Python-based sidecar microservice running alongside .NET Core 8 applications, which performs local LLM inference and semantic enrichment. The approach uses a... Read more