In today’s computing landscape, the most important breakthroughs no longer come from software alone or hardware alone, but from the increasingly powerful space where the two meet. Few engineers operate comfortably in that intersection. Sohil Grandhi is one of them.
His expertise is anchored by a rare dual-master’s foundation, a Master of Science in Computer Science and Engineering from Pennsylvania State University and a Master of Science in Machine Learning and Artificial Intelligence from LJMU, UK. This combination allows him to navigate the complexities of traditional systems architecture and the nuances of modern AI with equal fluency.
From validating next-generation silicon at Intel, to building AI-powered modem systems at Hughes Network Systems, Sohil’s career reflects a rare fusion of systems engineering, artificial intelligence, and large-scale data science. His work lives where theory meets deployment, where algorithms must survive contact with hardware, and where performance is measured not in lines of code, but in nanoseconds, watts, and real-world reliability.
Rather than following a single technical track, Sohil has spent his career deliberately crossing boundaries between disciplines. That cross-disciplinary mindset now defines his work in both industry and research.
From silicon to software: a foundation built at Intel
Sohil’s professional story began deep inside one of the world’s most complex computing environments: Intel’s silicon validation and hardware interconnect teams. Working on CPU-to-FPGA coherence using cutting-edge interconnect technologies such as CXL and UPI, he was responsible for ensuring that some of the most advanced processors in the world behaved exactly as designed under real operating conditions.
This was not theoretical verification. These systems powered real customers, real data centers, and real devices. Failures could mean product delays or multi-million-dollar consequences.
To handle that complexity, Sohil designed Python-driven automation frameworks that replaced slow, manual testing pipelines. These tools validated coherence transactions , monitored signal and power behavior, and verified performance across thousands of test cases. His innovations reduced testing time dramatically while improving reliability, saving significant operational cost.
He also developed energy-efficient voltage selection algorithms that optimized power usage in silicon without sacrificing stability, delivering up to 20% improvements in power efficiency during validation.
These innovations earned him multiple recognition awards and culminated in a high-profile Intel hardware demonstration, “Coherent Attach FPGA with XEON Processor,” which won the Viewer’s Choice Award at Intel’s System Hardware Summit.
That work laid the foundation for everything that followed: a deep, practical understanding of how real computing systems behave under pressure.
Bringing artificial intelligence into hardware
Rather than remaining purely in hardware validation, Sohil began asking a bigger question:
What if machine learning could make hardware systems smarter, faster, and more resilient?
That question became the focus of his graduate research and later his industry work.
At Hughes Network Systems, he applied deep learning directly to satellite communications, using CNNs, RNNs, and LSTMs to analyze modem signals and improve digital signal processing performance. These models were not confined to academic notebooks. They were deployed on Xilinx FPGA hardware, optimized through quantization and pruning, and integrated into live satellite systems via RESTful APIs.
The result was a 15% improvement in signal quality under real-world constraints, an achievement that required both machine learning expertise and hardware-aware engineering.
This blending of AI with embedded systems would later appear in Sohil’s academic work, particularly in his research on AI-driven cache coherence verification, where Graph Neural Networks were used to automate and accelerate one of the most difficult problems in modern chip design.
A research portfolio that mirrors real-world impact
Sohil’s academic work reflects the same philosophy that defines his engineering career: solve problems that matter in production systems.
His peer-reviewed research spans IoT, cloud computing, hardware verification, and big data, including:
- An IoT-based smart lighting system that used real-time occupancy data to optimize energy usage
- A novel GNN architecture that automated cache coherence verification for System-on-Chip designs
- An analysis of enterprise data evolution, mapping the shift from MapReduce to modern cloud-native architectures.
These works have been published in international journals and indexed by major academic institutions. His earlier book chapters on big data analytics and social network mining, published by Taylor & Francis and IGI Global, are held by university libraries including Stanford and Arizona State University.
Once again, his work lives at the boundary between code and hardware, where milliseconds matter and AI workloads push systems to their limits.
The engineer who bridges worlds
What makes Sohil Grandhi stand out is not any single credential, publication, or job title. It is his ability to connect domains that are usually siloed:
- Hardware validation with machine learning
- Cloud data architectures with embedded systems
- Academic research with production-grade engineering
Whether he is designing FPGA-friendly neural networks, automating verification flows, or publishing research on cache coherence, Sohil’s work is guided by one principle:
technology should be both intelligent and dependable.
In a world increasingly built on AI, high-performance computing, and massive data flows, engineers who understand only one layer of the stack will struggle. Sohil belongs to a smaller, more powerful group, those who understand how everything fits together.
And that is exactly where the future of computing is being built.
