Green leaf




Gaurav B - Deep learning developer
Member since:Dec 12, 2022
Profile last updated:Dec 06, 2022
Last activity:Jan. 1, 2023, 4:02 p.m. UTC
Location:
Canada
LD Interview score: 4.9 / 5
View Details

Gaurav B

I am a deep-learning researcher with an experience in computer vision, natural language processing, and data sciences. Currently, I am working as a P.h.D. researcher in the Machine Learning and Computer Vision Lab at the University of British Columbia. Previously I worked as Research Scientist at Descript-Inc, and as a Research Assistant at IIT-Hyderabad. My experience ranges from data analysis and statistics to implementing and deploying various machine learning models on Slurm and AWS.

Skills:    ·   ·   ·   ·   ·   ·   ·    ·    ·    ·    · 
Weekly Availability: 37 hours
DayUTCPST
Sun16 - 2208 - 14
Mon00 - 0516 - 21
Tues00 - 0516 - 21
Wed00 - 0516 - 21
Thurs00 - 0516 - 21
Fri00 - 0516 - 21
Sat16 - 2208 - 14

Hourly Rate:$31.25
Experience:7 yrs
Deep learning:6+ yrs
Computer vision:6 yrs
Machine learning:6 yrs
Natural language processing:4+ yrs
PyTorch:3+ yrs
Tensorflow:3+ yrs
keras:3+ yrs
Engineer's Devices:
Mac
Android
iPad

Experiences

University of British Columbia     Domain Translation with Deep Generative Models     Student     Employment
Sep 2021 - Dec 2025
Research

  • Developed a new deep generative model (a.k.a GAN) for solving domain translation problems in computer vision. The framework takes an image as input and generates a new image in another domain.
  • This framework can be used to translate the image of a male to the image of a female. The machine learning model can be used to convert images taken in daylight to night or taken in summer to winter.
  • The deep learning architecture is developed in PyTorch using Python. It is highly scalable and can handle multiple domains without adding more parameters. Thus the GPU requirement is very less.
  • This work also resulted in a research paper that has been published in WACV'23.
  • Publications - https://scholar.google.co.in/citations?user=PcmMT-4AAAAJ&hl=en

Skills used: Computer vision, Machine learning, PyTorch, Python, Deep learning



Link to the github:
Indian Institute of Technology Hyderabad     Deep generative and probabilistic models for Seq2seq learning     Developer     Employment
May 2020 - May 2021
Research

  • Created framework in Python using Pytorch that contains the implementation of deep probabilistic graphical models such as Variational Autoencoders (VAEs), Mixture Density Networks (MDN), and GANs.
  • The framework consists of attention-based seq2seq deep learning models.
  • The framework can be used for 3D sketch generation (computer vision) where the deep learning models complete an incomplete sketch automatically.
  • The framework can also be used to convert text into handwriting. The handwriting is personalized to each person by fine-tuning the model on a small handwriting dataset of the person.
  • It also supports machine translations in natural language processing which translates text from one language into another such as English-to-French, English-to-German, French-to-English, etc.

Skills used: Computer vision, Probabilistic graphical models, Natural language processing, PyTorch, Deep learning



Link to the github:
Descript-Inc     Speech generation and recognition     Developer     Employment
Jun 2019 - May 2020
Technology

  • Created the framework for speech recognition and generation using Python, Keras, and Pytorch. The framework was also used in the deployment of machine learning and deep learning models.
  • The framework consists of the implementation of deep learning models such as CNN, LSTMs, etc for audio synthesis, multi-class speech recognition, and tagging.
  • Used Amazon AWS and Google GCP for implementation of the framework.
  • This work resulted in a research paper on multi-classification and audio tagging in InterSpeech'18.

Skills used: Speech recognition, Machine learning, PyTorch, Python, Deep learning



Link to the github:
Indian Institute of Technology Roorkee     DeepLearn: Reproducing Deep Learning Papers on NLP and Data Sciences     Developer     Passion Project
May 2016 - Jun 2019
Research

  • Created an open-source framework for Natural Language Processing and Computer Vision tasks. The framework is written in Python using Keras, Scikit-learn, and Tensorflow, and is available on GitHub.
  • Implemented 15+ deep learning models from various research papers for many NLP and data sciences tasks. The deep learning models include attention-based CNN/LSTMs/Transformers.
  • This repository supports a ranking-based question-answer system. For a text question, the machine learning system produces top-K relevant answers based on contextual similarity.
  • It supports fake-news stance detection that can be used to classify whether a given news article is fake or not. It also supports finding documents written by a single author.
  • This repository also contains implementations of my research papers published in WWW'18, Pattern Recognition'19, and ACPR'17. Papers - https://scholar.google.co.in/citations?user=PcmMT-4AAAAJ&hl=en

Skills used: Computer vision, Machine learning, Natural language processing, scikit-learn, Tensorflow, Deep learning, keras



Link to the github:

Education

University of British Columbia    PhD  (Computer Science)
Aug 2021 - Aug 2025
Indian Institute of Technology Roorkee    Mtech  (Computer Science)
Aug 2015 - Aug 2017

Courses

Machine Learning    University of British Columbia
Jan 2021 - Jun 2022
Skills learnt: Probabilistic graphical models, Deep learning, Artificial intelligence, Machine learning
Probabilistic Programming    University of British Columbia
Sep 2021 - Dec 2021
Skills learnt: Probabilistic graphical models, Deep learning, Compilers, Machine learning
Visual Geometry    University of British Columbia
Sep 2021 - Dec 2021
Skills learnt: Deep learning, PyTorch, Computer vision