Experience

1910 Genetics

• United States
• Biotechnology Research
• 1 - 100 Employee

• Artificial Intelligence Research Scientist I

  • Feb 2023 - Present

  In my role as an Artificial Intelligence Research Scientist, I am dedicated to staying current with the latest advancements in artificial intelligence applied to drug discovery, molecular property prediction, docking methodologies, and 3D molecular generation. My work involves executing data curation, training, and testing pipelines for structure-based drug design. In my role as an Artificial Intelligence Research Scientist, I am dedicated to staying current with the latest advancements in artificial intelligence applied to drug discovery, molecular property prediction, docking methodologies, and 3D molecular generation. My work involves executing data curation, training, and testing pipelines for structure-based drug design.



Northwestern University - The Feinberg School of Medicine

• United States
• Higher Education
• 700 & Above Employee

• Postdoctoral Research Associate

  • Jul 2022 - Feb 2023

  As a Postdoctoral Research Associate at Northwestern University, I honed my skills in deep learning and medical imaging, specifically focusing on the diagnosis of aortic flow abnormalities using 4-dimensional flow MRI for cardiac applications. My work includes the development of two innovative deep learning models. The first is a CycleGAN, implemented with Tensorflow and Keras in Python, designed to predict 3D flow-MRI-derived aortic hemodynamics using aortic anatomy obtained from CT angiography. The second model I developed is a Mixture Deep Neural Network (DNN) comprising a Multi-Layer Perceptron (MLP) and a CNN, intended to diagnose aortic valve-induced abnormal flows. This model uses the scalograms of chest vibrations measured by seismocardiography. In addition to these technical achievements, my role involved collaboration with a multidisciplinary team of researchers, clinicians, and engineers. I also contributed to the dissemination of our findings through presentations and publications. This combination of technical innovation and teamwork has been crucial in advancing our understanding and diagnosis of aortic flow abnormalities. Show less



Stony Brook University

• United States
• Higher Education
• 700 & Above Employee

• Postdoctoral Research Associate

  • May 2022 - Aug 2022

  As a Postdoctoral Research Associate at Stony Brook University, my primary research focused on leveraging deep neural networks and terahertz spectroscopy to improve the prediction of histological markers in burn injuries, such as the rate of apoptosis and re-epithelialization. To achieve this, I developed a robust neural network model using Python, which significantly enhanced our ability to assess burn severity accurately.In addition to this, I implemented a nonlinear optimization technique in MATLAB, designed to extract Debye parameters from terahertz reflectivity. This approach allowed us to classify burn injuries more effectively as superficial, partial-thickness, or full-thickness. This innovative method not only improved the assessment of burn injuries but also led to the development of physics-based feature extraction techniques. These techniques were crucial in processing large datasets from clinical trials and enhanced the interpretability of black-box machine learning models.Apart from my technical contributions, I have worked collaboratively with a multidisciplinary team of clinicians, engineers, and fellow researchers. I have contributed to several publications and presentations based on our findings and have also played a key role in mentoring junior researchers in our lab. This balance of technical innovation, teamwork, and mentorship has led to substantial advancements in the field of burn injury assessment. Show less

• Graduate Research Assistant

  • Aug 2017 - May 2022

  As a Graduate Research Associate at Stony Brook University, I concentrated my research efforts on improving the triage of in vivo burn injuries. I did this by developing a range of machine learning and signal processing techniques. I created and implemented supervised machine learning models using support vector machines, random forests, and boosted linear discriminant analysis algorithms. This resulted in significant improvements in the diagnosis of burn severity using terahertz reflection spectra.To enhance the accuracy, sensitivity, and specificity in predicting burn healing outcomes, I leveraged the Shannon entropy of wavelet packet coefficients of terahertz time-domain waveforms. This novel feature extraction and dimensionality reduction approach revolutionized our predictive capabilities. In parallel to the development of machine learning models, I formulated a signal processing algorithm that improved the signal-to-noise ratio of in vivo terahertz spectra using level-based wavelet hard thresholding and Wiener deconvolution.Beyond these significant technical advancements, I contributed to hardware development by participating in the design and creation of a high-speed handheld terahertz scanner. By implementing sparse deconvolution to remove Gaussian light wave aberrations, I significantly improved the scanner's imaging resolution. My research also extended to studying the effects of rough surfaces and volume scattering on terahertz reflection spectra. This exploration led to the development of a novel broadband spectroscopic technique. We named this method the "bimodality coefficient spectrum of terahertz reflectivity". This innovative approach leverages higher-order statistics and wavelet multiresolution analysis in the Fourier domain to detect terahertz characteristic absorption lines even in the presence of severe scattering. Show less



Amirkabir University of Technology - Tehran Polytechnic

• Iran
• Higher Education
• 700 & Above Employee

• Undergraduate Research Assistant

  • Mar 2015 - Feb 2016

  As an Undergraduate Research Assistant at Amirkabir University of Technology, I dedicated my efforts to a project focused on enhancing the diagnostic accuracy of brain tumors. This was accomplished through the utilization of magnetic resonance spectroscopy (MRS), a highly sophisticated, non-invasive technique that allows for in-depth exploration of molecular composition. I implemented a signal processing approach based on the Stockwell transform to obtain the time-frequency distribution of MRS measurements of brain tumors. This proved instrumental in presenting an improved perspective on the intricate spectral details of the disease. Moving to the next stage of the project, I employed a Bayesian neural network to execute binary classification of the brain tumors. This classification was based on the previously derived Stockwell transform of MR spectra. Significantly, my research demonstrated that the technique I employed presented a higher level of accuracy than alternative techniques, such as the 2D wavelet transform or the Wigner-Ville distribution. This finding has important implications for improving the diagnostic reliability in the field of neuro-oncology, potentially impacting patient treatment outcomes and prognosis. Show less



Pardis Communications Company

• Tehran, Tehran Province, Iran

• Summer Intern

  • May 2015 - Sep 2015

  I worked with a team of electrical engineers to design and implement classical analog signal modulation and demodulation techniques, such as amplitude modulation, frequency modulation, and single- and double-side band modulations. I wrote all the programs in MATLAB. I worked with a team of electrical engineers to design and implement classical analog signal modulation and demodulation techniques, such as amplitude modulation, frequency modulation, and single- and double-side band modulations. I wrote all the programs in MATLAB.