Mohan Guo

About Me

I am a master's student in Artificial Intelligence at the University of Amsterdam, currently conducting my thesis at the Amsterdam Machine Learning Lab (AMLab) under the supervision of Prof. Patrick Forré and Cong Liu. My research focuses on frame-based equivariant diffusion models for 3D molecular generation. More broadly, I am fascinated by modeling structure in 3D space—spanning molecular modeling, 3D reconstruction, and physics-informed generative models.

I’m particularly interested in how geometric structure, symmetry, uncertainty, and representation learning can be combined to develop generative models that are expressive, data-efficient, and physically grounded. My goal is to build systems that respect the underlying inductive biases of 3D data—whether in molecular conformations or reconstructed scenes—and that generalize reliably to new, complex environments.

Outside of research, I’m drawn to the structure and serenity of nature—its hidden symmetries, emergent behavior, and the way simple principles give rise to complex forms. I enjoy hiking, photography, and reflecting on how natural phenomena translate into abstract patterns—something that resonates deeply with how I approach AI. The background image on this page was taken at Rakwa Tso in Tibet, one of my favorite places. Feel free to ask me about it if you're curious. :)

Projects

Frame-based Equivariant Diffusion Models

MSc Thesis · AMLab, University of Amsterdam · Jan – Jul 2025

Developed a geometric transformer framework for 3D molecular generation under the supervision of Prof. Patrick Forré and Cong Liu. Achieved -137.97 NLL (SOTA), 50% inference speedup, and 89.39% structural validity through systematic optimization of equivariant diffusion models.

AI Agent System Development

Internship · HAVAS Media Amsterdam · Nov 2024 – May 2025

Built a hallucination-resistant enterprise AI agent integrating domain knowledge graphs, symbolic reasoning, and uncertainty-aware validation. Delivered 95%+ factual accuracy in client-facing outputs; led stakeholder presentations to bridge AI solutions with business strategy.

E(n) Equivariant Diffusion Model Optimization

Research Project · University of Amsterdam · Apr – May 2024

Implemented JAX-based equivariant diffusion models with flow matching for 3D molecule generation. Demonstrated 60% performance improvement and improved synthetic accessibility prediction; published at GRaM Workshop @ ICML 2024.

Neural-symbolic Model in Causality Alignment

Research Project · University of Amsterdam · Jun – Aug 2024

Designed a hybrid neural-symbolic framework for causal representation learning in embodied AI, supervised by Prof. Sara Magliacane. Integrated symbolic constraints into loss functions to improve causal alignment in visual reasoning tasks (iThor environment).

Reproducibility Study of Equal Improvability Fairness

Research Project · University of Amsterdam · Feb – May 2024

Extended and validated an existing fairness framework. Investigated limitations under non-Gaussian assumptions and proposed bias mitigation strategies based on robust statistical modeling.

Real-time FMCW Radar Systems

ASIC Design Intern · Purple Mountain Laboratory · Sep 2021 – Jun 2023

Delivered a production-ready radar processor achieving 12.67ms latency, 55.65mW power, and 256-target tracking. Led end-to-end development from MATLAB simulation to FPGA prototyping and chip tape-out; results published in IEEE Access.

Publications

• Frame-based Equivariant Diffusion Model for 3D Molecular Generation.
  Mohan Guo*, Cong Liu*, Patrick Forré.
  NeurIPS 2025 Workshop on Foundations of Probabilistic Inference (FPI).
  [arXiv Extended Version] (* equal contribution)
• Equivariant Diffusion for Molecule Generation in 3D using Consistency Models.
  Sedlacek, M., Vozikis, A., Bartak, P., Cadigan, L., & Guo, M. In GRaM Workshop, ICML 2024.
• K/Ka-Band Hybrid-Packaged Four-Element Four-Beam Phased-Array Transmitter and Receiver Front-Ends with Optimized Beamforming Passive Networks.
  Yang, M., Zhao, D., Xu, C., Gu, P., Yi, Y., Guo, M., Yan, X., Liu, L., Chai, Y., Liu, H., & You, X. IEEE Journal of Solid-State Circuits, 59(11), 3142–3155. (2024)
• K/Ka-Band 4-Element 4-Beam Hybrid Phased-Array Transmitter and Receiver Front-Ends with Compact Layout Floor-Plans and Fault-Tolerant Digital Circuits.
  Yang, M., Xu, C., Gu, P., Yi, Y., Guo, M., Liu, L., Yan, X., You, X., & Zhao, D. In 2023 IEEE Asian Solid-State Circuits Conference (A-SSCC). (2023)
• An Integrated Real-Time FMCW Radar Baseband Processor in 40-nm CMOS.
  Guo, M., Zhao, D., Wu, Q., Wu, J., Li, D., & Zhang, P. IEEE Access, 11, 36041–36051. (2023)