Biography
Tianyu is a Principal AI Research Scientist at Autodesk AI Lab, where he focuses on applying machine learning to optimize decision-making and control processes. His work primarily revolves around graph-structured data, developing scalable algorithms for deep reinforcement learning, and model alignment. He is particularly interested in fine-tuning and adapting models to real-world scenarios, ensuring reliable and consistent output, even with limited training data.
Tianyu's research spans various areas, including sensor fusion, simulation environment design and implementation, transfer learning, diversity-induced reinforcement learning, multi-agent reinforcement learning, and combinatorial optimization. His expertise across these domains enables him to approach complex problems from multiple perspectives.
During his Ph.D., Tianyu was recognized for his academic excellence with the Alberta Graduate Excellence Scholarship. His master's and Ph.D. theses were also nominated for the departmental graduate award in 2019 and 2023, including the Outstanding MSc Thesis Award.
What's New?
- January 2024I'll be joining the Autodesk AI research team led by Dr. Tonya Custis in March 2024!
- December 2023I passed my PhD final exam. Big shout out to my advisor, Prof. Omid Ardakanian for his invaluable guidance!
- June 2023Our eEnergy workshop paper got accepted!
- May 2023We presented a poster on UpperBound AI week.
- May 2023Our journal paper is accepted! Check out why we should learn suboptimal policies.
- April 2023Congrats Abilmansur for his first paper!
Experience
Education
Doctor of Philosophy
Department of Computing Science
Advisor: Dr. Omid Ardakanian
Thesis: Data-Enabled Optimization of Building Operations
★ Nominated for Departmental Outstanding PhD of Science Thesis Award
My research is inter-disciplinary and utilizes different techniques from machine learning, control systems and communication networks. It could produce innovative data-driven solutions for optimizing the energy consumption of buildings in the context of a smart city.
Master of Science
Department of Computing Science
Advisor: Dr. Omid Ardakanian
Thesis: Building Occupancy and Thermal Modelling in the Wild
★ Departmental Outstanding Master of Science Thesis Award Runner Up
Bachelor of Science with Honors
Department of Computing Science
Three ACM ICPC Regional medals.
Work
Senior AI Research Scientist
Autodesk Research
Leverage large language models (LLMs) and reinforcement learning (RL) to enhance design automation, precision, and efficiency. Develop RL-based fine-tuning strategies to align LLM outputs with engineering design intent, using CAD-specific reward functions such as stability and fully-constrained metrics. Explore methods for improving reasoning diversity and long-context understanding in LLMs to support complex design tasks. This research aims to bridge AI and human-centric design, enabling smarter, more reliable, and production-ready CAD tools.
Research Assistant
University of Alberta (Part-time)
Assists with academic research include but are not limited to editing and preparing manuscripts, producing academic papers, performing research work in archives, analyzing data, and developing models.
Teaching Assistant
University of Alberta (Part-time)
- CMPUT 379 — Operating System Concepts
- CMPUT 366 — Reinforcement Learning
- CMPUT 272 — Formal Systems and Logic
Library Back-End Programmer
University of Alberta (Part-time)
Deal with university publications. This includes organizing files stored on OpenStack Swift and MySQL entries, structuring and uploading items to the university production server using Ruby on Rails. Uploading items to the Internet Archive.
Publications
Refereed Papers
- Evan Casey, Tianyu Zhang, Shu Ishida, John Roger Thompson, Amir Khasahmadi, Joseph George Lambourne, Pradeep Kumar Jayaraman, Karl D.D. Willis, “Aligning Constraint Generation with Design Intent in Parametric CAD”, In Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV), IEEE2025.
- Amin Rakhsha, Tianyu Zhang, Kanika Madan, Amir-massoud Farahmand, Amir Khasahmadi, “Majority of the Bests: Improving Best-of-N via Bootstrapping”, In Proceedings of the Second AI for MATH Workshop at the 42nd International Conference on Machine Learning (ICML)2025.
- Jing Yu, Tianyu Zhang, Omid Ardakanian, Adam Wierman, “Online Comfort-Constrained HVAC Control via Feature Transfer”, In Proceedings of the 16th ACM International Conference on Future Energy Systems (e-Energy), ACM2025.
- Tianyu Zhang, Omid Ardakanian, “Comfort-aware Optimal Space Planning in Shared Workspaces”, In Proceedings of the 15th ACM International Conference on Future Energy Systems (e-Energy), ACM2024.
- Tianyu Zhang, Omid Ardakanian, “Investigating the Impact of Space Allocation Strategy on Energy-Comfort Trade-off in Office Buildings”, In Companion Proceedings of the 14th ACM International Conference on Future Energy Systems (e-Energy), pp.145-149, ACMJune 2023.
- Abilmansur Zhumabekov, Daniel May, Tianyu Zhang, Aakash Krishna, Omid Ardakanian, Matthew E. Taylor, “Ensembling Diverse Policies Improves Generalizability of Reinforcement Learning Algorithms in Continuous Control Tasks”, In Proceedings of the Adaptive and Learning Agents Workshop (ALA 2023), 9 pages, ACMMay 2023.
- Aakash Krishna GS*, Tianyu Zhang*, Omid Ardakanian, Matthew E. Taylor, “Mitigating an Adoption Barrier of Reinforcement Learning-Based Control Strategies in Buildings”, Energy and Buildings, vol.285, 112878, ElsevierApril 2023.[Impact Factor: 7.201]
- Tianyu Zhang, Aakash Krishna GS, Mohammad Afshari, Petr Musilek, Matthew E. Taylor, Omid Ardakanian, “Diversity for Transfer in Learning-based Control of Buildings”, In Proceedings of the 13th ACM International Conference on Future Energy Systems (e-Energy), pp.556-564, ACMJune 2022.
- Tianyu Zhang, Amin Banitalebi-Dehkordi, Yong Zhang, “Deep Reinforcement Learning for Exact Combinatorial Optimization: Learning to Branch”, In Proceedings of the 26th International Conference on Pattern Recognition (ICPR), pp.3105-3111, IEEEAugust 2022.
- Tianyu Zhang, Jake Gu, Omid Ardakanian, Joyce Kim, “Addressing Data Inadequacy Challenges in Personal Comfort Models by Combining Pretrained Comfort Models”, Energy and Buildings, vol.264, 112068, ElsevierJune 2022.[Impact Factor: 7.201]
- Tianyu Zhang*, Gaby Baasch*, Omid Ardakanian, Ralph Evins, “On the Joint Control of Multiple Building Systems with Reinforcement Learning”, In Proceedings of the 12th ACM International Conference on Future Energy Systems (e-Energy), pp.60-72, ACMJune 2021.
- Md Monir Hossain*, Tianyu Zhang*, Omid Ardakanian, “Identifying grey-box thermal models with Bayesian neural networks”, Energy and Buildings, vol.238, pp.1-11, Elsevier2021.[Impact Factor: 7.201]
- Tianyu Zhang, Omid Ardakanian, “Poster Abstract: COBS: COmprehensive Building Simulator”, In Proceedings of the 7th ACM Conference on Systems for Built Environments (BuildSys), ACMNovember 2020.[Best Poster Runner Up]
- Md Monir Hossain*, Tianyu Zhang*, Omid Ardakanian, “Evaluating the Feasibility of Reusing Pre-trained Thermal Models in the Residential Sector”, In Proceedings of the 1st ACM International Workshop on Urban Building Energy Sensing, Controls, Big Data Analysis, and Visualization (UrbSys), pp.23-32, ACMNovember 2019.
- Tianyu Zhang, Abdullah Al Zishan, Omid Ardakanian, “ODToolkit: Extensible Toolkit for Occupancy Detection in Smart Buildings”, In Proceedings of the Tenth ACM International Conference on Future Energy Systems (e-Energy), pp.35-46, ACMJune 2019.
- Tianyu Zhang, Omid Ardakanian, “A Domain Adaptation Technique for Occupancy Estimation in Smart Buildings”, In Proceedings of the International Conference on Internet of Things Design and Implementation (IoTDI), pp.148-159, ACMApril 2019.
Preprints
- Md Monir Hossain*, Tianyu Zhang*, Omid Ardakanian, “Identifying Grey-box Thermal Models with Bayesian Neural Networks”, preprint available on arXiv (last revised in September 2020)
Theses
- Tianyu Zhang, “Data-Enabled Optimization of Building Operations”, Ph.D. Dissertation, University of AlbertaDecember 2023.
- Tianyu Zhang, “Building Occupancy and Thermal Modelling in the Wild”, M.Sc. Dissertation, University of AlbertaAugust 2019.
Presentations
- “Modeling and Simulation Studies and Tools: Co-simulation Platform for Occupant-centric Control of Buildings”, 6th International Symposium on Occupant BehaviourSeptember 2020.
People
My supreme supervisor
Omid Ardakanian
Associate Professor at the University of Alberta
My beloved girlfriend
Yang Liu
Account Manager Small Business at TD
My MVP ex-colleague
Aakash Krishna G.S.
Machine Learning Engineer at Resolver Inc.
My MVP colleague
Evan Casey
Researcher at Autodesk / Founder of Cadmium AI