CS Compiler Seminar: Jianming Tong, "Leveraging AI ASICs for Cryptography Acceleration."
Feb 19, 2026 4:00 - 5:00 pm
4403 Siebel Center
- Sponsor
- Compilers, Architecture, and Parallel Computing Research Area
- Speaker
- Jianming Tong
- Contact
- Allison Mette
- agk@illinois.edu
- Views
- 8
- Abstract: Artificial Intelligence (AI) is driving a new industrial revolution, transforming human workflows increasingly into digital tokens. However, this transformation exposes sensitive data at an unprecedented scale, leading to significant privacy challenges that have slowed AI adoption. Homomorphic Encryption (HE) offers strong data privacy guarantees for cloud services but comes with prohibitive computational overhead.
While GPUs have emerged as a practical platform for accelerating HE, there remains an order-of-magnitude energy-efficiency gap compared to specialized (but costly) HE ASICs. This talk explores an alternate direction: leveraging existing AI accelerators, such as Google’s TPUs, to accelerate homomorphic encryption and broader cryptographic primitives.
The focus is on advanced compilation techniques that transform applications with static scheduling of modular arithmetic into kernels natively supported by AI ASICs (e.g., TPUs) without hardware modifications. The CROSS project achieves state-of-the-art throughput for NTT and HE operators, as well as state-of-the-art energy efficiency among commodity devices including CPUs, GPUs, and FPGAs. This work demonstrates that HE operators can be transformed into TPU-suitable kernels, inheriting the energy efficiency and throughput of modern AI ASICs without hardware changes—opening a new direction for hardware-friendly cryptographic protocol design.Bio: Jianming Tong is a fifth-year PhD candidate at Georgia Tech, advised by Tushar Krishna. His research focuses on computer architecture for AI and cryptography, particularly enabling privacy-preserving AI systems without sacrificing performance. His work has been deployed at NVIDIA (NV Labs) and Google (Jaxite), and he has received several recognitions including 2nd place University Demo at DAC, the Qualcomm Innovation Fellowship, Machine Learning and Systems Rising Star, and the GT NEXT Award.