Jiajia Li
North Carolina State University, NC·Computer Science
Many hard computing problems, from large language models to quantum chemistry, reduce to operations on tensors: multi-dimensional arrays of numbers. Real data rarely fit these arrays neatly.
- Fields
Computer Science
Engineering
- Topics
high-performance computing
tensors
tensor networks
- Grants
- 3 known grants, through Jun 2029
What they work on
- Build a multi-level auto-tuning system that selects GPU execution strategies for ragged tensors, whose pieces vary in size, keeping overhead low via efficient sampling and search-space pruning.
- Develop communication and scheduling techniques improving data locality and load balance for irregular tensor workloads on single-device and multi-device GPU platforms.
- Create a compiler framework with adaptive operator fusion that integrates irregular tensor computations into existing AI systems, evaluated on language-model serving and heterogeneous graph workloads.
- Verify that mixed-precision tensor contraction sequences stay within error bounds using SMT solvers, automated logic checkers, with quantifier-free floating-point encodings and counterexample-guided abstraction refinement.
- Extend verification to contraction ordering and fusion through incremental SMT solving, portfolio solving and equality saturation, then identify implementations trading off FLOP count, memory and verified error.
- Design memory-heterogeneity-aware representations and balanced sparse tensor contraction algorithms for high-dimensional sparse tensor networks on systems mixing GPUs, TPUs and FPGAs.
- Cut the cost of sparse tensor networks using memoization, the reuse of earlier results, and intelligent allocation, and propose specialized accelerator architectures for these computations.
More about the research
Inputs vary in size, most entries may be zero (sparse), and speed-up tricks can quietly introduce numerical errors. These projects ask how to run irregular tensor computations quickly and correctly on modern hardware such as graphics processing units (GPUs), combining algorithms, compilers, computer architecture and formal verification.
Why it matters
More efficient handling of irregular tensors could make AI services such as large language models faster, cheaper and less energy-hungry. Verified tensor code could give chemists, physicists and materials scientists more confidence that optimized simulations are numerically correct. Efficient sparse tensor networks could improve compression and analysis of high-dimensional data in quantum circuit simulation, social networks and healthcare, with open-source tooling planned.
Known grants
From NIH and NSF research grants. Labs often have other funding we can't see.
Collaborative Research: SHF: Hierarchical Optimization of Ragged Tensor Operators for Deep Learning Workloads
NSF · Jul 2026 – Jun 2029
Collaborative Research: CS2: Formally Verified and Performance-Optimized Tensor Contraction Sequences in Quantum Many-Body Computations
NSF · May 2026 – Apr 2029
Collaborative Research: PPoSS: LARGE: Cross-layer Coordination and Optimization for Scalable and Sparse Tensor Networks (CROSS)
NSF · Sep 2023 – Aug 2028
Closest labs
Bo Yuan
Rutgers University-New Brunswick, NJ·Engineering·Early-career
Develop design principles for building domain expertise into learning-based robot motion planners, improving their reliability and performance in complex, resource-limited real-world settings.
Peng Jiang
University of Iowa, IA·Computer Science·Early-career
Develop a runtime system that gives a virtualized view of data partitioned across heterogeneous devices, plus a compiler tool that places data using a statistical performance model of random memory accesses.
Zachary Tatlock
University of Washington, WA·Computer Science
Speed up and extend the egg equality saturation library with relational e-matching (faster searches for places to apply rules), sketch-based extraction for choosing the best program, and destructive rewriting to control e-graph blow-up.
The school
North Carolina State University: 40 labs on PI Pages, $717M a year on research, 603 research doctorates a year.
Reach out
A start for your email. Fill in the brackets.