EMPLOYMENT History
|
|
|
Research Associate [ October 2020 – Present ]
|
|
Organization
:
|
University of Southern Queensland, Springfield Central,
Australia
|
|
Responsibilities:
|
· Develop Distress Inference System
from Audio Phone Calls using cutting edge Machine Learning and Artificial
Intelligence techniques.
|
|
Lecturer [ Semester 3, 2020-2021]
|
|
Organization
:
|
University of Southern
Queensland, Springfield Central, Australia
|
|
Responsibilities:
|
· I am appointed as an Assistant
Examiner in Machine Learning (CSC8003) and Big Data Management (CS8002),
where I will conduct Lecturing, Planning and Development, Course
Coordination, and Tutorial. I am also appointed to redevelop the CS8002
course with cutting edge knowledge.
|
|
Postgraduate Intern [ November 2020 – Present]
|
|
Organization
:
|
Data 61, CSIRO, Australia
|
|
Responsibilities:
|
· After joining CSIRO, I have
started to work on the Australian Coral Reef Monitoring System. Here, I am
developing a novel Artificial Intelligence model to "identify the
presence of coral from the video". Once I develop the model, it will
bring significant success to the coral reef monitoring system in Australia.
Therefore, we will be able to preserve the coral reef in Australia.
|
|
Research Scholar [ October
2017 – Present ]
|
|
Project
:
|
Advance Queensland Fellowship project to develop Mood
Inference Tool to predict the relapse in the mood disorder from audio phone
calls
|
|
Organization
:
|
University of Southern Queensland, Springfield Central,
Australia
|
|
Responsibilities:
|
· Collected and managed a large
amount of speech audio data and text data in the local development server.
Furthermore, conducted appropriate processing of the audio and text data
for feeding into machine learning models, which involved an enormous amount
of cleaning, transformation, and feature engineering
· Used cutting edge deep learning
models such as transformer, RNNs, seq2seq, CNN, etc. to build Sentiment
Classifier from text data and robust Speaker Identification system from
speech audio
· Built a system to detect four
affective events: anger, sadness, laughter, and cry from speech audio with
deep learning models using transfer learning techniques. I achieved a
classification accuracy of 98%, where state of the art is 72%. The demo can
be found on this link (https://youtu.be/E450z0ugCMs)
· Developed complex heuristic and
learning-based algorithm to detect the change of the mood of a user from
the phone conversation, maintaining the utmost privacy and security
· Collaborated with the Industrial
Partners; Netcare and Queensland Health to deploy a secured cloud-based
rest API serving Mood Inference Tool for end users
· Conducted extensive research on
Deep Neural Networks and proposed three novel models based on Generative
Adversarial Neural Network. These models aim to learn representation from
unlabeled audio data and generate high fidelity audio leveraging a small
amount of labelled data. Now, I am conducting research on few-shot learning
for audio data
|
|
Data Scientist [ November
2016 – October 2017
]
|
|
Organization
:
|
Infolytx, Dhaka, Bangladesh
|
|
Responsibilities:
|
· Developed a Human Activity
Detection System from a phone accelerometer and gyroscope using
bleeding-edge Deep Neural Networks. Furthermore, successfully collaborated
with the software engineering team to deploy this system on the cloud for
live detection for any smartphone device. The demo can be found here on https://bit.ly/37p0RTq
· Maintained test-driven
development for production codebase and followed agile methodology
· Built a Cervical Cancer Detection
System from 2D images of the sick patients where the underlying model was
built with CNN. This was a Kaggle Competition, and we were able to achieve
a prestigious rank in the competition
· Consulted the team to build a
robust commercial product identification system from images. The demo can
be found here on https://www.infolytx.com/solutions/deep-vision/
· Conducted a hands-on “Deep
Learning 101” course comprising of 27 lectures (2 hours each) for training
fellow team members
|
|
Software Engineer [ June 2015 – November 2016 ]
|
|
Organization
:
|
Data Robin, Khulna, Bangladesh
|
|
Responsibilities:
|
· Developed a fraud detection
system for a car insurance company and hosted the whole web service on a
website. After evaluating most of the traditional machine learning models,
XGBoost offered the highest accuracy of around 95%. For processing the written
description of the customers’, the pertained Word2vec and LSTM was used
· Built a unique tree-based neural
network that outperformed other states of the art Recurrent Neural Networks
for extracting business names from unstructured customer chat text
· Supervised a pilot project to
build a system to detect sick people from the 2d color image of a person
· Developed many exploratory data
analysis reports using advanced data visualization techniques
|
|
Mentor and Project Reviewer (Freelancing) [ August 2016 – Present ]
|
|
Courses
:
|
Data Science and Machine Learning
|
|
Organization
:
|
Udacity, California, USA (Remote)
|
|
Responsibilities:
|
· Teach and assist students in
machine learning and data science courses. As Udacity is an online
learning platform, students come from different parts of the world with
different backgrounds which, makes this job very challenging and rewarding
· Interact with the students so
that they can complete their Machine Learning/ Data Science Capstone
Projects successfully
· Evaluate Capstone Projects of the
students where many students come with their unique projects
|
|
SKill
|
|
|
Summary:
|
· Work with stakeholders to develop
a data-driven solution to bring a positive outcome for any business
· Data Wrangling, Database
Management, Exploratory Data Analysis, Data Visualization, Statistical
Analysis and Predictive Modelling using cutting-edge Machine Learning
models
· Deep Learning model building for
Computer Vision, Natural Language Processing, and Signal Processing
· GPU Computing for training and
deploying Deep Neural Networks
|
|
Programming
Languages:
|
Python, R, Bash, Java, JavaScript, C, C #, Matlab,
Octave, HTML, PHP, CSS, MySQL, Ajax, SQL, Scala, Node JS and Typescript
|
|
Deep
Learning Frameworks:
|
Tensorflow, Pytorch, Torch, Keras, Caffe, and Theano
|
|
Language:
|
Bangla, English, and Hindi
|
|
Others:
|
Docker, Git, Bitbucket Pipeline, Google Cloud Platform,
AWS environment, Cloud Formation, Linux Server Management, Serverless
Lambda, Flask, Django, AWS EC2, MySQL, Pyspark, MS Azure, Hadoop, Tableau,
Amazon Sagemaker, Numpy, Pandas, OpenCV, Scipy, Plotly, Matplotlib, D3js, and
MS Office stacks, Spark, Scala
|
|
Paper
|
|
· Nazmul
Haque, Kazi, Siddique Latif, and Rajib Rana. “Disentangled Representation
Learning with Information Maximizing Autoencoder.” arXiv preprint
arXiv:1904.08613 (2019).
· Haque,
Kazi Nazmul, Mohammad Abu Yousuf, and Rajib Rana. “Image denoising and
restoration with CNN-LSTM Encoder Decoder with Direct Attention.” arXiv
preprint arXiv:1801.05141 (2018).
· Haque,
Kazi Nazmul, Rajib Rana, and Björn Schuller. “Guided generative adversarial
neural network for representation learning and high fidelity audio generation
using fewer labelled audio data.” arXiv preprint arXiv:2003.02836 (2020).
[Under Review at IEEE/ACM Transactions on Audio, Speech, and Language
Processing]
· K.
N. Haque, R. Rana and B. W. Schuller, "High-Fidelity Audio Generation
and Representation Learning With Guided Adversarial Autoencoder," in
IEEE Access, vol. 8, pp. 223509-223528, 2020, doi:
10.1109/ACCESS.2020.3040797.
|
