Software Developer | Machine Learning

Work with progressive organisation that gives me the opportunity to utilise my skills in achieving common goals of the corporation and a bright personal career. Currently enrolled at the University of Queensland for Master of Data Science. Three and a half years of work experience in a highly scalable messaging platform and digital health web services. Passionate about software|data engineering & areas of AI like deep learning, NLP & computer vision.


Master of Data Science

Data Science Capstone Project
Explainable Deep Learning for Coral Classification
Coral taxonomy is currently undergoing a ‘molecular revolution’ which is fundamentally altering our understanding of the systematics and evolution of reefs. Importantly, molecular phylogenetic (tree of life) data is showing that micromorphological and microstructural features of the coral skeleton better reflect species evolutionary relationships than traditional macromorphological traits. While the combination of molecular phylogenetics and morphological analysis has facilitated better understanding of coral diversity, analysing microstructural features under the microscope is time-consuming and requires considerable expertise. Furthermore, this integrated taxonomic approach cannot be applied to the tens of thousands of coral specimens in the Queensland Museums (QM) collections collected before the time tissue preservation became necessary.
Machine learning could help overcome these problems by identifying and classifying species based on their skeletal morphology from high-resolution images, providing a valuable tool for coral taxonomy. Importantly, machine-learning tools could then be applied to accurately identify the extensive historical collections at QM, a task that would otherwise be completely unfeasible given the time required to examine each specimen.
The aim of the project is to build deep learning tools that can classify images of coral specimens into a hierarchical taxonomy also providing explanations for the picked category (e.g., using Methods include the use of computer vision models (e.g., pretrained imagenet models or training models from available coral image datasets).

Course's Semester 1
  • Operation Research & Mathematical Planning
  • Advanced Database Systems
  • Introduction to Data Science
  • Mathematics for Data Science
  • Course's Semester 2
  • Responsible Data Science
  • Machine Learning for Data Science
  • Data Mining
  • Applied Probability & Statistics
  • Course's Semester 3
  • Social Media Analytics
  • Data Analytics at Scale
  • Statistical Methods for Data Science
  • Data Science Capstone Project - Propose
  • Course's Semester 4

  • GPA: 5.875/7

    Summer Research Scholar | The University of Queensland

    Research Topic: Artificial intelligence for the prediction and prevention of concussion
    Used Mask RCNN framework to track players trained on the Cityscape Dataset and transfer learning. Accomplished detection of human pose estimation in video gameplay from pre-trained coco human pose estimation model. Implemented team tracking using OpenCV. Extracted video frames of tackles/concussions parsing from XML data. Detected collision using intersection over union(IoU).
    Supervisor: Dr Shakes Chandra

    CyberBullying Detection in Social Media | Toxic Comment Classification Challenge | Kaggle
    Cyberbullying is a crime where one person becomes the target of harassment, racism, toxicity and hate etc. A sequential deep learning architecture: Embedding, Convolution, Max Pooling, Dense layers eliminates the need for feature engineering and produces better prediction than traditional machine learning approaches using the concept of word embedding using a high-level API of TensorFlow open-source library.

    Analyzing the Factors Affecting Delivery Time | Olist | Kaggle
    E-commerce has increasingly become more popular as well as customer's expectations. Olist, an ecommerce platform, does not always predict an accurate delivery time. We go through the data science process to reduce the error in estimating the delivery time. After explorative data analysis, we identify the features that contributes to the longer estimated delivery time. In consideration to the findings of the analysis we make recommendations that can be considered to improve delivery experience.

    Software Developer | Zs Solutions Ltd

    Aglie Development
  • Responsive and agile development professional experienced in full project lifecycles. Deliver user stories for server-side application on a weekly basis or fortnight that is robust, functional, and scalable. Use SCRUM for agile development and participate in team-led solutions, reviewing peer’s code for quality and completeness. Identify development issues, brainstorm solutions, and constantly improve the solution via testing and feedback until the issue is resolved. Write unit tests and proper documentation of code. Deploy code to test servers and production
  • Experience
  • Worked for SmartIdeas group of companies delivering innovative products, services and solutions for Digital Health and Integrated Messaging on a global scale
  • Lead developer for web application registration module with payment gateway. Accomplished upgrading Python EOL 2.7 to 3.7 and Django Framework 1.55 to 2.2 LTS for the application. LifeMed
  • Contributed to a highly scalable messaging platform and web services, built using Tornado asynchronous server, Celery task scheduler, SQLAlchemy the ORM. Contributed to a Custom Messaging Service product, Parcelforce Worldwide. Lead Developer for the implementation of two-factor authentication, password reset & necessary RESTFUL-APIs in order to comply with GDPR. Improved report generation performance. SmartMessage
  • Undergraduate Degree

    Bachelor of technology in computer science & technology from National Institute of Technology, Rourkela, India
    Final year thesis on the study of an item based collaborative filtering, recommender System
    Graduated with first-class grade