CV
Jie Tian's academic curriculum vitae, research, teaching, and technical background.
LinkedIn · Google Scholar · GitHub
Education
University of Nevada, Reno — Ph.D. student in Mathematics with a specialization in Applied Mathematics, August 2024–Present. Expected graduation: May 2028. GPA: 3.930/4.0.
Completed coursework and passed qualifying exams in Real/Functional Analysis, Linear/Numerical Linear Algebra, and Nonlinear Dynamics with High Pass. Dissertation research is ongoing.
Shanghai University — M.Sc. in Pure Mathematics, September 2021–June 2024. Average: 91.92/100.
Hebei University of Science and Technology — B.Sc. in Mathematics and Applied Mathematics, September 2017–June 2021. GPA: 3.8/4.0.
Research Interests
- Matrix theory, quaternion algebra, and tensor computation
- Numerical and randomized linear algebra, low-rank approximation, and representation learning
- Convex optimization and computational methods on manifolds
- Inverse problems and uncertainty quantification
- Stochastic processes, discrete-time and continuous-time Markov chains, and time-series analysis
- Machine learning, LLM and agent post-training, knowledge distillation, and reinforcement learning
Research Experience
LLM and Agent Post-Training and AI Safety
LLM and Agent Post-Training Research, University of Nevada, Reno — September 2026–Present
- Independently design and implement experiments in knowledge distillation and LLM post-training, including supervised fine-tuning (SFT), reinforcement learning with verifiable rewards (RLVR), on-policy distillation (OPD), and self-distillation.
- Develop PyTorch/Transformers training and evaluation workflows with student rollout generation, frozen-teacher scoring, sampled-token teacher–student rewards, and clipped on-policy updates.
- Build reproducible experiment controllers for multi-GPU veRL configurations, including Hydra settings, checkpointing, validation schedules, run metadata, and log parsing.
- Develop diagnostics for GPU memory, training failures, reward/KL, top-k overlap, teacher–student consistency, entropy, gradients, timing, throughput, and task accuracy; extend studies toward agentic and function-calling settings.
- Investigate multi-teacher on-policy distillation and privacy risks through membership inference evaluation.
Mathematical and Computational Research
Department of Mathematics and Statistics, University of Nevada, Reno — August 2024–Present
Quaternionic Upper Bild Convexity and Optimization
Study the convexity and geometric structure of quaternionic numerical ranges through constrained real optimization and manifold methods. Develop boundary characterizations and computational validation for generic and degenerate matrix cases. Preprint with Tin-Yau Tam: Revisiting the Upper Bild of Quaternionic Numerical Ranges (2026).
FFT-Based Quaternion Tensor Decomposition, PCA, and ICA
Develop FFT-based quaternion tensor decompositions and spectral methods for image and video data, preserving coupled color-channel structure. Investigate principal component analysis (PCA), independent component analysis (ICA), and multilinear discriminant analysis for dimensionality reduction and representation learning.
Open-Source Contributions
EffLoRA: Memory-Efficient PiSSA Initialization for Hugging Face PEFT
- Implemented
replace_lora_weights_pissafor layer-wise PiSSA initialization of models already loaded in 4-bit. PR #4135 merged into Hugging Face PEFT’smainbranch in November 2026. - In the public PR benchmark on Qwen3-8B, peak GPU allocation during initialization decreased from 15.9 GiB to 6.4 GiB (about 60%) compared with the documented PEFT workflow, with bitwise-identical outputs in the tested configuration.
Related talk abstract submitted to the 2027 SIAM Conference on Computational Science and Engineering (CSE27).
Selected Course Projects
Delay-Induced Oscillations in a Logistic–Allee Model with Harvesting — Math 721, December 2025. Project report (PDF).
Analyzed stability and Hopf bifurcation using Erlang-chain approximations; derived characteristic equations, computed critical delays, and studied convergence of the approximation.
Modeling and R₀ Analysis of SEQIRS and SEIR–SEI Epidemic Models via the Generalized Linear Chain Trick — Math 722, May 2026. Project report (PDF).
Modeled phase-type disease-stage durations and derived the basic reproduction number using next-generation matrices, examining quarantine and host–vector transmission.
These projects support ongoing study of discrete-time and continuous-time Markov chains, transition matrices, generators, and absorption times, including connections between continuous-time Markov chains and Erlang/phase-type distributed delays.
Publications
Preprint: Jie Tian and Tin-Yau Tam, “Revisiting the Upper Bild of Quaternionic Numerical Ranges” (2026). arXiv:2611.04044 · DOI.
Coauthor of three peer-reviewed papers in Symmetry (2022), Computational and Applied Mathematics (2024), and Mathematics (2024). See the publication list for full citations.
Presentations
- Qt-Eigenvalue Theory of Third-Order Quaternion Tensors — Joint Mathematics Meetings (JMM 2026), Washington, DC, January 5, 2026. Includes numerical experiments on Qt-PCA-SD. Slides (PDF).
- Generalized Eigenvalues of Third-Order Quaternion Tensors and Multilinear Discriminant Analysis Under Qt-Product — SIAM Annual Meeting (AN26), Cleveland, OH, July 6–10, 2026. Slides (PDF).
See Talks for the full list, including invited talks at JMM 2025 and ILAS 2026.
Upcoming invited talk: “So–Thompson’s Proof of Upper Bild Convexity via Convex Optimization,” ILAS–JMM Special Session on Matrix Analysis and Applications, Joint Mathematics Meetings, Chicago, January 12–15, 2027.
Teaching and Service
University of Nevada, Reno — Graduate Teaching Assistant / Instructor of Record, August 2024–Present
Fall 2026 — Instructor of Record, Math 120E: Fundamentals of College Mathematics (Expanded) (4 credits; 4 instructional hours per week).
Independently develop course materials and lesson plans, design quizzes and examinations, assess student work, and provide academic support. Topics include sets and logic, probability and statistics, consumer mathematics, variation, geometry and trigonometry for measurement, and linear, quadratic, exponential, and logarithmic functions, emphasizing problem solving and applications.
Previous teaching includes Calculus I and II and Precalculus II. See Teaching.
Shanghai University — Teaching Assistant, September 2021–June 2023.
Vice President, SIAM Student Chapter, UNR — January–May 2026. Organized academic lectures and research seminars and supported interdisciplinary networking.
Technical Skills
- Mathematical foundation: Convex optimization, numerical and randomized linear algebra, and computational methods on manifolds.
- Programming: Python (PyTorch, Transformers, NumPy, SciPy, pandas), MATLAB, R, basic C; Git, Linux, and LaTeX.
- Research infrastructure: veRL, vLLM, FSDP, Ray, Hydra, multi-GPU experiment configuration, checkpointing, experiment logging, monitoring, and training diagnostics.
- Methods: Quaternion matrix/tensor algorithms, FFT, PCA/ICA, LoRA, stochastic processes, Markov chains, and phase-type distributions.
Certifications
- IBM Data Science Professional Certificate
- IBM AI Engineering Professional Certificate
- IBM Generative AI Engineering Professional Certificate
Scholarships and Selected Awards
- Aydin Arman Ozbek Ph.D. Endowment — $500, Department of Mathematics and Statistics, University of Nevada, Reno (June 2026)
- Ben & Beatrice Edwards Math Award — $1,800, Department of Mathematics and Statistics, University of Nevada, Reno (June 2026)
- Outstanding Graduate Assistant Award — $500, College of Science, University of Nevada, Reno (May 2026)
- Second-class University Scholarship (Top 15%), Shanghai University (2022–2023)
- Song Yuqi Scholarship, College of Science, Shanghai University (2022)
- First-class University Scholarship (Top 10%), Shanghai University (2021–2022)
- National Scholarship, China (2018–2019)
- First-class Scholarship (Rank 1/33), Hebei University of Science and Technology (2019–2020)
Travel Grants and Conference Support
- SIAM Annual Meeting 2026: SIAM Travel Award, $650, Society for Industrial and Applied Mathematics.
- ILAS 2026: Departmental Travel Support, $750, University of Nevada, Reno; ILAS Early Career NSF Conference Grant Award, $300.
- Joint Mathematics Meetings 2026: Departmental Travel Support, $750, and Graduate Student Association Travel Grant, $500, University of Nevada, Reno.
- Joint Mathematics Meetings 2025: Departmental Travel Support, $750, and Graduate Student Association Travel Grant, $500, University of Nevada, Reno.
Professional Experience
Data Analyst Intern, Shanghai Jinghe Information Technology Co., Ltd. — January–August 2021
Conducted data scraping, cleaning, and statistical analysis to support business decisions.