I'm Javad Zolfaghari Bengar, a Senior AI/ML Engineer with a PhD in Computer Vision. I turn ambitious research into intelligent systems—from perception and industrial AI to next-generation wireless technologies.
Selected research and engineering work. Public-facing descriptions only; some implementation details are intentionally omitted.
Exploring phase-aware deep learning for wireless channel estimation, bridging model development with hardware-in-the-loop validation. Presented at IEEE CSCN 2026.
Developed and evaluated AI models for automated optical inspection of semiconductor defects, combining image and inspection metadata for defect severity assessment. Contributed to model validation and integration into an industrial inspection workflow. Proprietary data, implementation details and unpublished metrics are intentionally omitted.
Led the research and architecture of Evershelf, a video-based human action recognition system designed to identify potentially suspicious activity, including possible theft, in retail settings. Focused on translating advances in video understanding into a practical real-world AI application.
Computer vision for complex real-world waste streams, including detection, instance segmentation, and domain adaptation in industrial settings.
Restored original portfolio animation, optimized for fast loading.
Video-based active learning research with synthetic driving sequences and rich annotations, including segmentation, 2D/3D bounding boxes and depth. The publicly hosted train and test packages have been downloaded more than 119,000 times combined.
Restored original portfolio animation, optimized for fast loading.
Addressing biased sample selection in imbalanced datasets to make data-efficient image classification more effective.
Investigating how self-supervised representation learning can help reduce human annotation effort.
Portfolio content describes publicly shareable research and high-level professional contributions. Employer names indicate professional affiliation, not endorsement.
Sharing working AI systems with the research community and industry.
Contributed to and presented two AI-driven wireless technology demonstrations at Keysight's booth: digital twins for 6G and AI-based channel estimation, connecting applied research with practical test and measurement.
Our team received the Best Demo/Poster Award for Hardware-in-the-Loop Phase-Aware CNN for Real-Time 5G Channel Estimation at IEEE CSCN 2026.
Read the demonstration paper ↗Different domains. The same ambition: make machine learning useful beyond the lab.
Senior AI/ML Engineer · AI Labs, Barcelona
AI for wireless test and measurement and industrial vision; translating research concepts into evaluated, demonstrable engineering solutions.
Lead Researcher & System Architect · Evershelf
Led research and system architecture for video-based human action recognition to flag potentially suspicious retail behaviors, including possible theft. Described at a high level to respect employer confidentiality.
PhD in Computer Vision
Research on reducing annotation effort with deep active learning, video perception, learning dynamics, and self-supervised methods.
Peer-reviewed and publicly available work across computer vision and wireless AI.
My background spans applied mathematics, communications engineering, and a PhD in computer vision. Over the years, I've worked on problems that demand both research depth and engineering pragmatism: selecting the right data, building reliable models, and making them useful in the real world.
Today, I'm especially interested in the intersection of AI, physical systems, perception, and wireless technologies—and in taking technically ambitious projects from early ideas through real validation.
Open to thoughtful conversations about AI research, applied machine learning, and ambitious engineering challenges.
Start a conversation ↗