Md. Omur Faruque

Jr. Software Engineer at Dynamic Solution Innovators Ltd.

+8801857282578

Chittagong, Bangladesh

Profile image

NOTABLE PROJECTS


Home (Django, PostgreSQL)

A real estate web app for property listing where realtors can add properties and people can query with multiple filters.

Tweety (PHP, MySQL)

A twitter clone built using core PHP. Basic functionalities include twitting, commenting, sharing, chatting etc.

An encoder decoder architecture for image captioning with CNN encoder and LSTM decoder. Images were from COCO dataset. Pretrained InceptionV3 model was used to extract features for encoder via transfer learning.

Machine Translation (TensorFlow)

An English to German machine translation model using TensorFlow’s custom layers and custom training loops.

Chatbot (Reformer, Trax)

A Reformer model is trained on the MultiWoz Dialogue Dataset to create a chatbot. Model needs further tuning and training but still grasps the content of the dialogue and generates sentences with syntax preservation more often than not.

Duplicate Finder (Python, Trax)

A Siamese network model trained on Quora questions pair dataset which detects duplicates with an F1 score around 80%.

VAE for CelebA (TensorFlow Probability, VAE)

A Variational Autoencoder model is trained on the subset of CelebA dataset, which can reconstruct and generate new celebrity images. Latent space is also explored and changed in desired direction to change attributes of input images.

KitKat (Android, Socket)

An android P2P chatting app with basic chatting and text file sharing capabilities. Only works with users on the same WiFi network.

Spaceball (Unity, C#)

A 3D endless runner game.

About me


“I love solving puzzles, playing chess. I get fascinated with mathematics and the relentless growth of machines and technology.

Deeply into Neural Nets.”

Education


B.Sc. in Computer Science & Engineering

Shahjalal University of Science and Technology

Sylhet, Bangladesh

Skills


Python

C, C++

Java, C#

TensorFlow

Django

Numpy, Pandas

Matplotlib, Seaborn

MySQL, PostgreSQL

Git

Thesis


Sentimental Style Transfer in Text with Variational Autoencoder

  • A probabilistic model using VAE and holistic discriminator to transfer the style of texts while controlling the semantics. (i.e., converting a paragraph to Shakespearian style). Primarily worked with Positive-Negative sentiments.
  • Used Bengali language.

Others


  • Obtained 30+ certiifications and completed around 40 courses from different MOOC sites like Coursera, DataCamp, Udacity etc.
  • Solved 200+ problems on different online judges.