Fine-Tuning ProGen2 Protein Language Model
Fine-tuned ProGen2 transformer model on PFAM protein families to generate biologically valid sequences. Achieved 93%+ motif retention rates through targeted hyperparameter …
•
1 min read
🔬Ashkan Ghanavati
PhD Candidate | ML & Materials Science
Bridging experimental materials science and machine learning to solve complex, data-scarce problems. I specialize in extracting maximum insight from limited datasets through intelligent experimental design and physics-informed algorithms.
Carbon nanotube fiber assembly - Scanning Electron Microscopy
Technical Skills
Programming & ML
Python
PyTorch, TensorFlow, scikit-learn
Machine Learning
Supervised, Unsupervised, Deep Learning
MLOps
MLflow, Airflow, Docker, GCP
Experimental Skills
Materials Characterization
SEM, Raman, Mechanical Testing
Data Analysis
Feature Engineering, Visualization
Experimental Design
DOE, Bayesian Optimization
🎯 Research Focus
My research bridges experimental materials science and machine learning. I work with real, noisy experimental data from carbon nanotube fiber assemblies, applying techniques like Bayesian optimization and data-efficient ML strategies to extract maximum insight from limited datasets.
I’m passionate about solving problems where physical intuition meets computational power - where domain knowledge guides algorithms, and algorithms reveal patterns that experiments alone cannot uncover.
Featured Projects
Physics-Informed Neural Network for 2D Heat Equation
Built a physics-informed neural network in PyTorch that embeds the transient 2D heat equation PDE directly into the loss function, achieving high accuracy validated against …
•
1 min read
Machine Learning for Carbon Nanotube Fiber Characterization
Building intelligent experimental workflows for CNT fiber optimization by integrating comprehensive materials characterization with machine learning. Developed automated data …
•
1 min read
End-to-End MLOps Pipeline for Customer Segmentation
Built production-ready ML pipeline for customer segmentation with automated workflows, containerization, and cloud deployment. Demonstrates end-to-end MLOps practices from data …
•
1 min read
Experience
Graduate Research Assistant
Northeastern UniversityDeveloping novel CNT assemblies with ML-guided parameter optimization. Built automated Python pipelines for data collection and analysis.Graduate Teaching Assistant
Northeastern UniversityTeaching thermodynamics, fluid mechanics, MLOps, and mechanics of materials lab to 150+ students per semester.
Education
PhD Mechanical Engineering
Northeastern UniversityMSc Energy Conversion
University of TehranBSc Mechanical Engineering
University of Tehran