Leonid Popryho

About

I am a Ph.D. student at the University of Illinois Chicago, advised by Prof. Inna Partin-Vaisband, graduating in May 2028. I work on machine learning for chip design in two directions. Generative models for physical design: diffusion and reinforcement learning that propose routes and Steiner trees directly, instead of searching for them. And physics-informed surrogates: graph neural networks that learn from PDE solvers and then answer design questions in a fraction of a second, with active learning deciding which expensive simulations are worth running.

Before the Ph.D. I was a machine learning engineer at Receptor.AI in Kyiv, working on drug-target affinity prediction and diffusion models for molecular docking. In both cases the data are graphs, molecules then and chips now, and the physics is what makes the problems hard.

I am looking for summer 2027 internships developing machine learning models for electronic design automation and chip design. LinkedIn is the best way to reach me.

News

Sep 2026: Preprint of Mesh-Native Physics-Informed Graph Surrogates, our ICCAD 2026 paper on TCAD-in-the-loop design exploration, is on arXiv.
Jul 2026: Mesh-Native Physics-Informed Graph Surrogates for TCAD-in-the-loop design space exploration accepted to ICCAD 2026.
Jul 2026: Electro-thermal optimization framework for TSV networks accepted to IEEE TCAD.
Jun 2026: Preprint of SHIFT, a compute-relocation framework for chiplet-based systems, submitted to IEEE.
May 2026: PALTO, physics-informed active learning for FinFET power delivery, presented by A. Sadeghi at IEEE ECTC 2026.
Mar 2026: Preprint of Electro-thermal optimization framework for TSV networks submitted to IEEE TCAD.
Mar 2026: Paper and talk on accelerating Sentaurus device exploration with physics-based active learning at SNUG Silicon Valley 2026.
Aug 2025: Preprint of RF-informed graph neural networks for circuit performance prediction is on arXiv.
Aug 2023: Started the Ph.D. at the University of Illinois Chicago.

Education

2023 - 2028: Ph.D., Electrical and Computer Engineering, University of Illinois Chicago, HiPerCAS Lab. Advisor: Inna Partin-Vaisband. Expected May 2028.
2022 - 2023: M.Sc., Computer Science, Blekinge Institute of Technology, Karlskrona, Sweden. Dual degree with Kyiv Academic University through Erasmus+. Advisor: Oleksandr Adamov.
2021 - 2023: M.Sc., Computer Science, Kyiv Academic University, Kyiv, Ukraine. Advisor: Nataliia Kussul.
2017 - 2021: B.Sc., Applied Mathematics, Igor Sikorsky Kyiv Polytechnic Institute. July 2021.

Experience

Research

May 2026 - Aug 2026: Research Aide, X-ray Science Division, Argonne National Laboratory, Lemont, IL.
AI-enabled co-design for low-temperature electronics: physics-informed neural network and graph surrogate models that replace coupled thermal-electromagnetic simulation, with sim-to-sim transfer across solver fidelities.
Aug 2023 - present: Graduate Research Assistant, HiPerCAS Lab, University of Illinois Chicago.
Machine learning for chip design, advised by Prof. Inna Partin-Vaisband: diffusion and reinforcement learning for physical design (global routing, rectilinear Steiner trees), physics-informed graph surrogates for circuit, interconnect and device simulation, and active learning over TCAD.

Industry

Sep 2024 - Jan 2025: Hardware Technology Intern, Apple, Cary, NC.
Machine learning for real-time on-device inference in C firmware.
Feb 2021 - Aug 2023: Machine Learning Engineer, Receptor.ai, Kyiv, Ukraine.
AI for drug discovery, end to end: from CI/CD, infrastructure and cloud AI services (AWS SageMaker, GCP Vertex AI) to chemical property prediction, molecule clustering, generative models for small molecules and drug-target affinity prediction; co-authored two RSC Advances papers.

Publications Google Scholar · 96 citations

Mesh-Native Physics-Informed Graph Surrogates for TCAD-in-the-Loop Design Space Exploration
L. Popryho, A. Sadeghi, I. Partin-Vaisband
IEEE/ACM International Conference on Computer-Aided Design (ICCAD), 2026 [arXiv] [PDF]
From Physics to Surrogate Intelligence: A Unified Electro-Thermo-Optimization Framework for TSV Networks
M. Gharib*, L. Popryho*, I. Partin-Vaisband (*equal contribution)
IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems (TCAD), 2026, early access [DOI] [IEEE] [arXiv] [PDF]
PALTO: Physics-Informed Active Learning for Tri-Gate FinFET Design Optimization for Vertical Power Delivery
A. Sadeghi, L. Popryho, I. Partin-Vaisband
IEEE 76th Electronic Components and Technology Conference (ECTC), 2026 [IEEE] [arXiv] [PDF]
Accelerating Sentaurus Device Exploration via Physics-based Active Learning: Case Study in GaN FinFET Optimization
L. Popryho, A. Sadeghi, I. Partin-Vaisband
SNUG Silicon Valley 2026 (Synopsys Users Group), paper and presentation [session] [proceedings]
SHIFT: Dynamic Compute Relocation Framework for Communication-Aware Chiplet-Based Systems
A. Delavari, L. Popryho, S. Swaroopa, N. Sehatbakhsh, I. Partin-Vaisband, B. Vaisband
arXiv preprint, submitted to IEEE, 2026 [arXiv] [PDF]
RF-Informed Graph Neural Networks for Accurate and Data-Efficient Circuit Performance Prediction
A. Asadi, L. Popryho, I. Partin-Vaisband
arXiv preprint, under review at IEEE, 2025 [arXiv] [PDF]
Augmenting a training dataset of the generative diffusion model for molecular docking with artificial binding pockets
T. Voitsitskyi, V. Bdzhola, R. Stratiichuk, I. Koleiev, Z. Ostrovsky, V. Vozniak, I. Khropachov, P. Henitsoi, L. Popryho, R. Zhytar, S. Yesylevskyy, A. Nafiiev, S. Starosyla
RSC Advances, 14, 1341-1353, 2024 [DOI]
3DProtDTA: a deep learning model for drug-target affinity prediction based on residue-level protein graphs
T. Voitsitskyi, R. Stratiichuk, I. Koleiev, L. Popryho, Z. Ostrovsky, P. Henitsoi, I. Khropachov, V. Vozniak, R. Zhytar, D. Nechepurenko, S. Yesylevskyy, A. Nafiiev, S. Starosyla
RSC Advances, 13, 10261-10272, 2023 [DOI] [code]

Conferences

45th IEEE/ACM International Conference on Computer-Aided Design (ICCAD 2026) · San Jose, CA · November 2026
Presenting “Mesh-Native Physics-Informed Graph Surrogates for TCAD-in-the-Loop Design Space Exploration.”
63rd Design Automation Conference (DAC 2026) · Long Beach, CA
DAC Young Fellows poster: “One-Shot Steiner Tree Prediction with a Graph cVAE.”
Synopsys Users Group, Silicon Valley (SNUG 2026) · Santa Clara, CA
Paper and presentation: “Accelerating Sentaurus Device Exploration via Physics-based Active Learning.”
62nd Design Automation Conference (DAC 2025) · San Francisco, CA
DAC Young Fellows poster: “Active-Learning-Based Device Optimization.”
61st Design Automation Conference (DAC 2024) · San Francisco, CA
Poster: “Multi-Terminal Pathfinding with Conditional Denoising Diffusion Probabilistic Model.”
Midwest Machine Learning Symposium (MMLS 2024) · Minneapolis, MN
Poster: “Multi-Terminal Pathfinding with Conditional Denoising Diffusion Probabilistic Model.”

Honors and awards

2025, 2026: DAC Young Fellow, Design Automation Conference. Includes a $500 travel grant in 2026.
2022 - 2023: Erasmus+ scholarship, funded dual-degree study at Blekinge Institute of Technology, Sweden.
2020: Kaggle “Where am I?” competition: 2nd of 25 teams.

Teaching

Spring 2024 Spring 2026: Teaching Assistant, ECE 350: Principles of Automatic Control, University of Illinois Chicago.
MATLAB/Simulink labs on transfer functions, state space, stability, feedback and digital control.
Spring 2025 Fall 2026: Teaching Assistant, ECE 333: Computer Communication Networks, University of Illinois Chicago.
Wireshark labs (TCP/IP, DNS, HTTP); web server, UDP pinger and SMTP projects.

Projects

Radar-camera 3D object detection (2025).
Course project for CS 518, Deep Learning for Computer Vision, advised by Prof. Sathya Ravi: how to fuse radar and camera data. Built on HGSFusion and the View-of-Delft benchmark. The main contribution is an ablation that evaluates the impact of each fused branch separately, alongside experiments with backbone fine-tuning, radar-occupancy-gated cross-attention fusion, regularization and class-aware anchor tuning, and exploratory analysis of the radar point cloud. PyTorch, OpenPCDet, CUDA.
Blackout notifier (2022 - present).
Telethon-based bot that monitors electricity status and alerts residents about scheduled power outages before the lights go out. Running since 2022 and still in use; has notified 500+ users. The chart compares actual outages with the published schedule, hour by hour, for one week.
Explosion crater detection (2022).
Mapping explosion craters from the war in Ukraine in free Sentinel-2 imagery. At 10 m resolution a crater is only a few pixels wide, so I framed it as per-pixel segmentation: large scenes and hand-labelled masks are tiled into 64×64 multispectral patches and a compact U-Net in PyTorch labels every pixel. It finds about 92% of labelled craters on held-out tiles, and 92% of what it flags is a real crater. The lesson was choosing the metric: a first model scored 85% pixel accuracy while detecting nothing at all. Advised by Prof. Nataliia Kussul. See the pipeline and results summary.

Misc

Pickleball. First place in the competitive division of my local Fourth of July tournament, July 4, 2026. Happy to lose to you on a court in Chicago.
Skydiving. Jumped out of a plane at 14,000 feet over Chicago in 2024. Tandem, but the view was all mine.
Kaggle. Placed 2nd of 25 teams in the “Where am I?” GameLevel competition in 2020.
Roots. From Kyiv; in Chicago since August 2023. Before that I lived in Sweden on an Erasmus+ scholarship, finishing a master’s degree. Several projects above exist because the war made them necessary.