Ishan Rai

Programmer

Location
Indian Institute of Technology, Roorkee, India
Email
Website
https://ishanrai05.github.io
GitHub
ishanrai05
LinkedIn
ishanrai

Experience

–

Google Summer of Code at CEPH Foundation

Highlights

  • Enhanced Ceph Dashboard UI using Angular to improved community branding.
  • Implement a style guide to give the UI and documented a process for designing and deciding mockups and designs for Dashboard.

–

Google Summer of Code at CERN-HSF

GANGA (Gaudi/Athena and Grid Alliance) is an interface used by scientists to interface with huge amount of computing power and storage available to them as part of the LHC computing grid.

Highlights

  • Created a cached memory management system in GANGA by implementing a flyweight design pattern using copy-on-write principle.
  • Programmed monitoring tools using cProfiler and memory-profile and audited Memory and CPU consumption
  • Wrote unit tests to reduce development timings and bugs

–

Open Source Contributor at Chainer

Chainer is an open source deep learning framework written purely in Python on top of NumPy and CuPy Python libraries.

Highlights

  • Implemented Numpy and deep learning routines ChainerX in C++ and CUDA. Wrote language bindings for python (using pybind11)
  • Simplified tests for more than 35 functions including activation, array, math and loss functions.
  • Overall 44 merged Pull Requests.

–

Deep Learning Internship at Indian Institute of Science (IISc)

Highlights

  • Coded an Encoder, BiRNN and BiLSTM aggregator and Decoder architecture in Tensorflow.
  • Created subroutines to pre-process images using Numpy and OpenCV and custom loss functions such as SSIM and MSSIM.

–

Full Stack Web Developer Intern at Diginique Techlabs

Highlights

  • Developed a website using Django, JavaScript, HTML, CSS and PostgreSQL
  • Set up Python WSGI HTTP server using Gunicorn and reverse proxy using Nginx. Hosted website on Amazon Web Services EC2

– present

Web Developer at Watch Out! News Agency

Highlights

  • Involved in maintaining the website for official campus magazine. Worked in Jekyll, Bootstrap and SASS.
  • Performed Web Scraping to get all the details about placement for the college

Projects

Addressing Large Hadron Collider Challenges by Machine Learning ||

Highlights

  • Particle Identification: Trained Gradient Boosting Classifier and Keras based Neural Network Classifier to identify electrons, protons, muons, kaons, pions and ghost particles using data from different responses in the tracking system, ring-imaging Cherenkov detector (RICH), electromagnetic and hadron calorimeters, and muon system. Plotted ROC curves using matplotlib and signal efficiency dependence from particle momentum and transverse momentum values.
  • Detector optimization: Optimized parameters of geometry of a simple tracking system using Bayesian optimization with Gaussian processes, Grid Search and Random Search.
  • Search for electromagnetic showers: Designed XGBoost Classifier prediction model to discriminate the base-tracks belonging to the electromagnetic showers from the background base-tracks.
  • Z Boson Mass: Analysed the mass spectrum of particle decaying into muon-antimuon (dimuon) pairs from Double Muondataset at CERN opendata and figured out the mass of Z boson portal using scikit-optimise toolkit.
  • Search for Rare Decay: Designed AdaBoost Classifier prediction model using data collected by the LHCb detectors togive similar performance on Monte Carlo simulated data as well as real data without correlating with mass.

Bayesian Methods for Machine Learning ||

Highlights

  • Implemented an app for creating facial composite using Variational Autoencoders in Keras and Gaussian processes that construct desired faces without explicitly providing databases of templates. Used Bayesian optimization with GPyOpt to maximize similarity function between original features and features reconstructed from the current point of continuous latent space.
  • Used probabilistic programming library PyMC3 to perform approximate Bayesian inference for logistic regression to model how the probability of a person having a high salary is affected by their age, education, sex and other features. Performed the Markov Chain Monte Carlo method to find credible intervals. Used Hamiltonian Monte Carlo algorithm for sampling and Metropolis-Hastings algorithm for finding the samples from the posterior distribution.

Practical Reinforcement Learning ||

Highlights

  • Used Cross-Entropy method with Sklearn’s MLP Classifier based Neural Network to solve the CartPole and MountainCar env in OpenAI Gym.
  • Implemented a basic reinforce algorithm, policy gradient in Tensorflow for OpenAI CartPole env.
  • Built a Deep Reinforcement Learning Agent using Tensorflow for Atari Kung FuMaster and trained it using Advantage Actor-Critic on parallel gym environments.
  • Implemented a vanilla Monte Carlo Tree Search (MCTS) planning and used it to solve OpenAI Gym CartPole env.

Traffic Survey as a Service (TSaaS) || IIT Roorkee

Highlights

  • Developed a RESTful web application using Django, React and PostgreSQL to create traffic surveys.
  • Set up a reverse proxy server using Apache2.

Languages

English
Fluency: Native speaker
Hindi
Fluency: Native speaker

Skills

Programming Languages
Level: Master
Keywords:
  • Python
  • Java
  • C++
  • C
  • R
  • JavaScript
Data Science
Level: Master
Keywords:
  • Scikit-learn
  • Machine Learning
  • Deep Learning
  • Chainer
  • Tensorflow
  • Pytorch
  • Pandas
  • Numpy
  • Matplotlib
  • OpenCV
Web Development
Level: Master
Keywords:
  • Django
  • Javascript
  • Flask
  • Jekyll
  • ReactJS
  • SQL
  • Sass
  • HTML
  • CSS
  • Apache2
  • Nginx
  • Gunicorn
DevOps
Level: Master
Keywords:
  • Git
  • Docker
  • GitHub Actions
Others
Level: Master
Keywords:
  • Qiskit
  • Linux
  • Bash

Interests

Open Source
Keywords:
  • Frequent contributer to Open Source
Debating
Keywords:
  • Member at Debating Society
  • Winner: Elocuation Contest on National Handloom Day organised by Ministry of Human Resource and Development
  • Winner: Asian Parliamentary Debate, Thomso'18, IIT Roorkee
  • Runner Up: Inter Hostel Debate - IIT Roorkee, 2018

Awards

Google Summer of Code

Selected as 1 of 1199 out of 51244 students from 178 countries.

CERN Summer Student

Selected out of 2500 applicants to work in CERN, Meyrin, Switzerland for summers of 2020.

Google Summer of Code

Selected as 1 of 1276 out of 30922 students from 148 countries

JEE Advanced, 2017

Secured a rank within top 0.2 percentile among 0.22 million candidates

Bill Gates Award

from Sunbeam English School for being best in Computer Science

Education

– present

Bachelor in Civil Engineering from Indian Institure of Technology, Roorkee with 7.765/10 CGPA

Courses

    Intermediate from Sunbeam English School, Varanasi with 94.0%

    Courses

    • Physics
    • Maths
    • Chemistry
    • English
    • Computer Science

    High School from Casterbridge School, Ballia with 95.83%

    Courses

    • Maths
    • Science
    • Social Science
    • English
    • Computer Science
    • Hindi