Traditional Machine Vision relies on rigid rule-based algorithms. MLLMs like GPT-4V are changing the game by allowing zero-shot defect detection with simple natural language prompts...

Privacy and latency are barriers to cloud AI. With new accelerators like Jetson Orin and quantized "TinyLLMs", we can now run powerful inference completely offline...

AI-driven Generative Design explores thousands of permutations to create complex, organic structures optimized for Additive Manufacturing, shifting engineers to "Constraint Architects"...

Deep Learning provides perception; Symbolic AI provides reasoning. Combining them creates "Reliable AI" that can optimize processes while strictly adhering to safety constraints...

Robots are leaving the cage. Driven by Embodied AI foundation models, general-purpose humanoid robots are beginning to learn physics and tasks end-to-end...

Industrial robots have traditionally relied on rigid programming. In this post, I discuss my recent experiments using Deep Reinforcement Learning (DRL) to allow 6-axis arms to adapt to variable environments in real-time...

While C++ remains the king of embedded performance, Python is rapidly gaining ground in the OT layer. I compare latency, libraries, and development speed for SCADA backends...

Transitioning from pure hardware control to software-heavy data science wasn't easy. Here are the courses, projects, and mindset shifts that helped me bridge the gap...