Independant Projects

Deep Learning, IOT Acoustic Anger Detector (Ongoing Team Project)
  • Using crowd-labelled Emotion datasets to develop a well performant portable Anger Classifier using supervised deep learning using CNNs and AutoEncoders
  • Plan is to Run this classifier on an edge device such as Raspberry Pi 4 which can be placed
    • Using: Python, Keras with Tensorflow as backend, NodeRed, NodeJS
Deep Learning Bat Species Classifier (Ongoing Team Project)
  • Used supervised learning to develop a CNN based wellperforming classifier with an aaccuracy and F1 measure of upto 97%
  • Compared the impact of Melspec Inputs vs MFCC Inputs vs STFT Inputs on the perfomance (mainly F1 scores and Accuracy) of the models
  • Reported the above research in the form of a conference paper at the AI for Good Global Summit 2020. The conference paper got accepted as it so the future plan is to coninue further development mainly focusing on optimizing the model and reducing the size
  • Used: Python, Jupyter Notebook, Spyder, Keras with Tensorflow as backend
Linux and Bash Programming, Email Security DMARC Mail Check App
  • Used an AWS virtual machine (EC2) instance to intantiate a test mail server.
  • Used open-source packages and setup an SPF, DKIM and DMARC filter to secure the mail server
  • created monthly crone servicesto run xustomesed self-written bash scripts to store and retreive Mail Logs
  • Built an User Interface to view summary reports of DMARC compromized emails being filtered out
    • Functionality: On the app, the user can download complete reports or view summary statisitcs by choosing customized dates.
  • Used: Linux EC2 instance, HTML, Java, Javascript, CSS, MySQL.

Course Projects

Deep Learning and Neural Networks Emotion Classification Model
  • Learnt the process of Data Cleaning
  • Made use of important concepts of batch normalization,dropout layers, maxpooling, activation functions and CNNs for feature extraction and dense layers for classification
  • Analysed the F1- scores, Precision Recall Curves , ROC curves and Area under the curves to select the best model.
  • Used: Python, Jupyter Notebooks, Spyder, GPU RTX4000, Keras with Tensorflow as backend
Cloud Computing Smart and Secure Blood and Organ Managing Web app deployed on Microsoft Azure
  • Used virtual machines,subnetting, gateways, firewalls, and cosmos Data Base to create the app
  • Created a RESTful API that can be used by third-party applications
  • Designed a user Interface for direct usage.
  • Functionality: The app connects various blood and organ banks within a city allowing them to get information about availble organs and blood types to reduced time wastage during emergency situations.
  • Used: HTML, Java, SpringBoot, Spring MVC, Microsoft Azure
Software Engineering Food Donation Web App
  • Used an Incremental Software Development approach to execute this team project built using the Onion Architecture
  • Personally contributed extensively towards the application logic and Black Box Testing
  • Functionality: The app connects restaurants who have food in excess and wish to donate with individuals and organizations who need food. Restaurant supply information on the food they are willing to donate and those in need can reserve this food. The goal of this app is to reduce food wastage in society
  • Used: Java, Spring Boot, Spring Data, maven
Software design HyperJump (Game)
  • Functionality: The Player must jump when the led window moves to the lower end of the spherical loop to gain positive points.Any Jumps outside the scoring zone will lead to negative points.If the player is playing well the LED window size decreases in length as player has entered difficult mode.
  • Used: Java, Arduino