Experience

  1. Lead AI Research Scientist, Precision Oncotherapeutics AI Modeling Lead

    Natera Inc.

    (Ranked exceeding (top 10%) for 2 consecutive years)

    • Implemented a end-to-end Vision Transformer (ViT) pipline for neoantigen prioritization which predicts the HLA binding likelihood of a given neopeptide.
    • Developed a protein language model to predict post-translational modification likelihoods from short protein sequences.
    • Designed and developed a data pipeline and a survival prediction model to infer patient-specific somatic mutations associated with differential response to platinum-based therapies.
  2. Lead Statistician and Machine Learning Research Scientist, Head of MCED/ECD/tfMRD/COO Biomarker Discovery Algorithms

    Natera Inc.

    (Ranked exceeding (top 10%) for 2 consecutive years)

    • Designed and implemented CpGTools which is the software infrastructure and the computational pipeline of the newly established Early Cancer Detection (ECD) and tissue-free Minimal Residual Disease (tf-MRD) program which is now the basis for all computational methylation efforts at Natera.
    • Designed and implemented the first completely reference-free methylation biomarker discovery pipeline. (Patent Application Submitted)
    • Developed a novel Hierarchical Bariational Bayes model to infer differentially methylated biomarkers from deep whole genome bsulfite sequencing data. (Patent Application Submitted)
    • Implemented a transformer VAE to learn the whole genome methylation patterns of healthy cfDNA samples.
    • Designed Natera’s CRC ECD/tfMRD methylation biomarker panel. (Patent Application Submitted)
    • Formed and directed the Cell-Type of Origin Pursuit Group (C2pG). A team of 4 dedicated to the cancer cell of origin problem.
    • Designed Natera’s first MCED biomarker panel
  3. AI Lead Scientist

    Ravel Biotechnology Inc.

    Cancer Prediction Model, Seq2Seq modeling and motif discovery, Cell-type Decomposition

    • Developed deep Seq2Seq models to infer functional events, e.g. histone modifications from accessibility patterns in cfDNA and utilized attribution priors to discover highly predictive motifs from regularized attributions.
    • Developed a variational Bayes model to infer cell-type of origin from functional events e.g. methylation and histone modification.
  4. Senior Machine Learning Research Scientist

    Ravel Biotechnology Inc.

    Fragmentomics panel design, cancer prediction modeling

    • Designed Ravel’s fragmentomics targeted panel
    • Developed Ravel’s main high-dimensional low sample size cancer prediction framework.
    • Utilized integrated gradients to find cancer predictive DNA fragmentation patterns
  5. Post-Doctoral Associate

    Chan Zuckerberg BioHub Initiative (CZI)
    Developed Explainable AI models for Genomic Data
  6. Contract/Non-Profit Machine Learning \& Data Science Consultant

    Department of Statistics, University of California, Berkeley
    Provided consulting services on a variety of topics e.g. Deep Learning, Neuroscience, Linguistics, Public Health, International Development, Genomics, Bio-chemistry etc. to academic and industrial clients.
  7. Statistical Learning Research Associate

    Lawrence Berkeley National Laboratory
    Collaborated with the Molecular Eco-Systems Biology Division and the Daphnia Consortium in mining, modeling and interpretation of genomic data.
  8. Ensemble Machine Learning Research Intern

    Illumina Inc.
    Designed and implemented DEnsLe, a deep variational ensemble learner, in PyTorch, to construct a ensemble learner that models the dependence structure between multiple different SNV callers.
  9. Lead Data Analyst

    Genapsys Inc.
    • Implemented a MATLAB pipeline to perform QC of raw output signals of Genius™ short-read genome sequencers.
    • Statistical analysis of the sensor array network output, e.g. clustering, correlation analysis, and base-calling.

Education

  1. PhD Statistics (Emphasis on Optimization and Machine learning)

    University of California, Berkeley

    Large Scale Interpretable Multi-View Learning with Application To High-Dimensional Multi-Omics Advisors & Thesis Committee: Prof. Peter J. Bickel, Dr. James B. Brown, Prof. Gary Karpen & Prof. Haiyan Huang.

    • MuLe: Formulated and implemented a novel large-scale interpretable and stable multi-view learning method for high-dimensional datasets with low sample size and utilized it to infer co-regulated modules between Metabolomics, Trasciptomics, and Microbiomics data collected on fruitfly samples in a treatment/control experiment.
    • BLOCCS: Extended MuLe to perform block matrix decomposition leading to more interpretable orthogonal canonical covariates.
    • SparKLe: Formulated and implemented a novel multiple kernel learning approach applied to kernel dimensionality reduction, implemented in.
    • Implemented a spark implementation of the three previous packages on Spark-MLlib.
    Read Thesis
  2. MA Applied Mathematics

    University of California, Berkeley

    Thesis Advisers: Prof. Per-Olof Persson & Prof. Alberto Grunbaum.

    • Numerically efficient implementation and benchmarking of second-order PDE solutions, specifically the wave equation (Python and C++).
    • Advection-Corrected Correlation Image Velocimetry: Numerical Computation of the velocity fields of Saturn’s surface vortices from Cassini images.
  3. B.Sc. Computational Fluid Dynamics

    Sharif University of Technology
    Computational patricle image velocimetry, modeling the transfer and deposition patterns of aerosols in the upper airways via finite volume analysis in C++ and OpenFoam.
Skills & Hobbies
Technical Skills
Pytorch Ecosystem (Lightning, Geometric, Ray, Pyro, fastai, Captum)
Statistical Modeling and Inference (Multi-View Learning, High-Dimensional Learning, Variational Inference)
Hobbies
Welding (MIG, TIG, Stick, Spot)
Algorithmic Trading
Motorcycles
Dog-walking
Awards
Chan Zuckerberg Biohub Initiative Award
CZI ∙ August 2019
Sally & Terry Speed Award
UC Berkeley ∙ September 2018
UC Berkeley Excellence Award
University of California ∙ July 2012
UC Berkeley Fellowship for Graduate Studies
University of California ∙ Present
UC San Diego Fellowship
University of California ∙ Present
University of British Columbia Scholarship
University of British Columbia ∙ Present
Languages
100%
English
100%
Farsi
50%
Arabic
50%
French
25%
German