About Me
About me
Building AI where physics meets deep learning to accelerate automation for a true Industry 4.0.
About Me
I’m Pranav Krishna, a researcher at the intersection of artificial intelligence and industrial automation, driven by a singular mission: leveraging deep learning and AI to revolutionize Industry 4.0 and push the boundaries of what’s possible in intelligent systems.
My Research Philosophy
My passion for AI research stems from working at the intersection of theory and practice. Having spent time both developing deep learning models and implementing control systems in industrial environments, I’ve learned that the most interesting challenges often arise when these worlds collide. There’s something uniquely valuable about understanding how elegant mathematical concepts translate to systems that must operate reliably under real-world constraints—noisy sensors, safety requirements, and economic pressures that pure simulation can’t capture.
Bridging Theory and Reality
My master’s research at UC Davis exemplifies this philosophy. I developed hybrid machine learning thermal space models for predictive control, creating an innovative framework that combines first-principles physics with encoder-decoder LSTM networks and feedforward neural networks. This wasn’t just academic exercise—it was about solving real problems where traditional control theory hits its limits and where pure machine learning approaches fail due to physical constraints.
The breakthrough came in my approach to handling noisy, incomplete industrial data. I designed novel training methodologies for hybrid models that can learn from limited sensing while respecting fundamental physical laws. This work addresses one of the most critical challenges in deploying AI in industrial settings: making deep learning robust enough for mission-critical applications.
Unique Research Strengths
What sets me apart as an AI researcher is my deep understanding of the full stack—from low-level control systems and PLCs to high-level machine learning architectures. I’ve programmed industrial automation systems at Genentech, developed advanced control algorithms at Yokogawa, and optimized complex chemical processes at ONGC. This gives me insights into data generation, sensor limitations, and system constraints that pure AI researchers often lack.
My technical background includes Python, MATLAB, advanced neural network architectures, state estimation techniques, deep reinforcement learning and optimization algorithms. More importantly, I understand how these tools must adapt when deployed in real industrial environments with safety constraints, regulatory requirements, more importantly pharmaceutical regulations like GxP, and economic pressures.
Vision for AI and Industry 4.0
I believe we’re at an inflection point where AI will fundamentally transform industrial systems. My research vision centers on developing AI that doesn’t just optimize existing processes, but reimagines how intelligent systems can orchestrate complex industrial operations. This means:
- Creating hybrid AI systems that seamlessly integrate physics-based models with data-driven learning
- Pioneering new approaches to transfer learning that allow AI models trained in one industrial context to rapidly adapt to others
- Advancing multi-modal AI that can simultaneously process sensor data, visual information, and process parameters
The Future I'm Building
I’m drawn to the fundamental challenges that lie at the intersection of AI and physical systems. Through my work with hybrid models—combining physics-informed approaches with deep learning—I’ve seen how this fusion can address some of the toughest problems in AI deployment: robust learning under uncertainty, safe operation in critical systems, and creating AI that truly understands physical processes rather than just pattern matching.
I believe hybrid modeling represents a promising path forward because it leverages the best of both worlds: the interpretability and reliability of physics-based models with the adaptability and pattern recognition capabilities of neural networks. This approach feels particularly important as we work toward truly autonomous industrial systems that must operate safely in complex, real-world environments.
My goal is to collaborate with world-class researchers where I can push the boundaries of what’s possible when you combine deep learning with physics based first principle models. I believe the next breakthrough in AI won’t come from bigger models alone, but from AI systems that truly “understand” (using equations from fundamental heat, mass, and momentum transfer balances) and can manipulate the physical world.