Ph.D. Candidate, University of Southern California
I develop machine learning methods that accelerate scientific discovery by combining modern AI with domain expertise. My research focuses on developing AI for engineering and scientific discovery, with applications in analog and RF circuit design, lipid nanoparticle engineering, and reliable generative AI.
Advised by Prof. Salman Avestimehr.
Developing machine learning methods for engineering, biomedical, and scientific discovery.
Machine learning for circuit benchmarking, topology-aware modeling, inverse design, electromagnetic modeling, physical synthesis, and automated layout generation.
Learn MorePredictive, interpretable, and generative machine learning for lipid nanoparticle (LNP) transfection, biodistribution, and targeted therapeutic delivery.
Learn MoreTest-time steering and uncertainty-aware routing methods for improving the quality, reliability, and efficiency of generative AI systems.
Learn MoreSequence models and intelligent AI systems for financial forecasting, cryptocurrency markets, and algorithmic trading.
Learn More
FALCON: An ML Framework for Fully Automated Layout-Constrained Analog Circuit Design
NeurIPS 2025
A fully automated machine-learning framework for analog circuit design that selects topologies based on performance specifications and infers parameters using layout-aware graph neural networks.
A Machine Learning Benchmarking Framework for Lipid Nanoparticle Transfection Efficiency Prediction
Communications AI & Computing, Nature Portfolio, 2026
A comprehensive ML benchmarking framework for predicting LNP transfection efficiency, enabling systematic comparison of models and guiding therapeutic delivery design.
ATHENA: Adaptive Test-Time Steering for Improving Count Fidelity in Diffusion Models
ViSCALE Workshop @ CVPR 2026
A model-agnostic, training-free test-time steering framework that improves object count fidelity in text-to-image diffusion models by estimating counts during sampling.
Decoding Extrahepatic Targeting of Lipid Nanoparticles with Interpretable Machine Learning
2026
An interpretable machine-learning framework for predicting LNP biodistribution and guiding extrahepatic targeting strategies for therapeutic delivery.
Selected for the Capital One fellowship through USC’s Capital One Center for Responsible AI and Decision-Making in Finance (CREDIF).
Communications AI & Computing · Nature Portfolio · 2026
NeurIPS · 2025
CVPR ViSCALE Workshop · 2026
Manuscript in Preparation · 2026
ISCAS · 2026
* Equal contribution
USC Capital One Center for Responsible AI and Decision-Making in Finance (CREDIF)
Qualcomm
Machine Learning and the Physical Sciences Workshop @ NeurIPS
University of Southern California
I am open to research collaborations and opportunities involving AI for science, machine learning, and scientific design.
mehradfa@usc.edu