Milad Pourrahmani, Ph.D.

Research Scientist


Irvine, CA


 LinkedIn > miladiouss  GitHub > miladiouss StackExchange
                  > miladiouss Twitter > miladiouss Milad's UCI Webpage


University of California, Irvine

Doctor of Philosophy, Physics — 3.66 GPA Dec 2019
Ph.D. Thesis: Deep Mining and its Astronomical Applications
Master of Science, Physics — 3.66 GPA Dec 2019

Purdue University, Indianapolis

Bachelor of Science in Physics — 3.81 GPA Dec 2012

Hard Skills

  • Advanced Python Programming
  • Performance Optimization
  • Data Visualization
  • PyTorch
  • Keras
  • TensorFlow
  • Pandas, NumPy, SciPy, Scikit-learn
  • Linux
  • JavaScript / HTML / CSS
  • d3js
  • Mathematica

Soft Skills

Experienced Data Miner

Proficient in developing a data pipeline for large datasets, from data collection and curation to training and fine-tuning neural net architectures.

Artistic Data Visualizer

Designer of creative and inviting data visualizations to understand, communicate, solve, and debug problems and to present the face of a project.

Creative and Collaborative Project Designer and Manager

A professional in designing, executing, and leading projects, independently, with internal teams, or by collaborating with external groups.


  • Machine Learning
  • Statistical Analysis
  • Data Science
  • Computational Science
  • Algorithm Design
  • Data Visualization
  • NLP



Aug 2019-Present
Applications Developer at California Institute of Technology
Postdoctoral Researcher at University of California, Irvine in Collaboration with JPL and NASA
  • Lead scientist for quality assurance and engineering the pointing sequence of a NASA space telescope named SPHEREx to optimize mapping of the (deep) sky, in a collaborative effort with 70 scientists from JPL, CalTech, and ASU.
  • Developed a toolkit for measuring the performance of the survey design algorithm by constructing a multi-dimensional map of the sky from a sequence of telescope pointings, involving large data processing on distributed CPUs and GPUs with optimized Python algorithms.
  • Large and high dimensional data analysis and visualization in Python and JS that is being used to debug and optimize the survey algorithm to ensure the requirements from NASA headquarters are met.
  • Conducting and maintaining hundreds of case studies to ensure best science quality.
  • Leading a project to designing a survey algorithm optimizing the observing time in the deep sky regions by real-time decision-making.
  • Our collaboration has led to $250 M of funding from NASA.

Deep Learning Image Classification on Astronomical Data with 50 Training Examples

Sep 2015-Aug 2019
Graduate Researcher at University of California, Irvine
  • Designed, tested, and published a new deep CNN architecture with PyTorch for biased and limited datasets (less than 50 examples) to identify rare astronomical events known as gravitational lenses among millions of common galaxy images.
    Ph.D. Thesis: Deep Mining and its Astronomical Applications
  • Mentored graduate and undergraduate students for various projects with leveraged interpersonal and communicational skills and by advocating critical and independent thinking.

Image Processing Pip Package

Personal Project
  • ezimage, a library published on pip repository to facilitate loading, reformating, and displaying images in Python.
  • Smart input recognition to load images from URLs, local paths, or data in different formats without any specifications.
  • Useful for machine learning to convert between CHW and HWC format for PyTorch and TensorFlow.

Using a Generative Adversarial Network (GAN) to Synthesize Gravitational Lenses

  • Used for boosting the training examples for image classification.
  • A generator was constructed from a series of transposed convolutions to upsample a latent vector to an RGB image.
  • A discriminator was used to identify synthesised images from real images.
  • A loss function was assigned to be maximized by the generator and minimized by the discriminator.
  • The discriminator was re-trained on synthesised data from 20 differently initialized GANs and used for data mining.

A New Kind fo Chess (Link)

Summer 2018
Wolfram Summer School
  • Created a symbolic representation of a chess game in the Wolfram Language.
  • Developed a chess board visualization.
  • The program outputs the list of all the possible moves as "rules" for the search algorithm.
  • Approved by Stephen Wolfram to be integrated as part of the Wolfram|Alpha website.

Student/Tutor Matching App Design

  • Recruited a group of undergraduate and master student software developers.
  • Set up a project development environment for the group.
  • Structured the backend and frontend and communicated its design to the developers.
  • Applied for grant funding and pitched the demo to entrepreneurs.

Awards and Honors

Honorary fellow in UC-Irvine Machine Learning and Physical Sciences program 2017-2018
Graduate Assistance in Areas of National Need (GAANN) 2017-2019
First Place Poster Presentation Award, UCI Dept. of Phys. & Astronomy at Research Presentation 2018
Scholar's List Recipient for Outstanding Academic Performance from School of Science Fall 2011 & 2012
Outstanding Learning Assistance from AAPT 2012
Student Employee of the Year for Outstanding Performance Spring 2012