Recent Work
Deep Learning Based Method To Understand Signaling Dynamics In Intestinal Stem Cells
Leveraging deep learning, the project achieves reliable and precise detection and tracking of
intestinal stem cell (ISC) trajectories across video frames to study signalling dynamics.
A commendable mAP50 value of 90.73% is obtained by implementing YOLOv8-M for cell detection and integrating
with a ByteTrack Multiple Object Tracking Precision of 78.47%. This approach underscores the potential of
deep learning to unravel temporal kinase dynamics within ISCs.
View More
Exploring the Dynamics of Startup Growth in Greece: A Comprehensive Data Analysis for the Data Science Hackathon 2023
Project involves the analysis of startups in Greece for a Data Science hackathon.
The analysis includes regional clustering, equity analysis, and sentiment analysis of tweets.
It can be useful for investors, entrepreneurs, and policymakers in making informed decisions,
such as identifying startups worth investing in, designing policies to support startup growth,
and identifying industries and regions with high potential for growth.
View More
Developing Analytical and Forecasting Solutions to Enhance Recruitment at HC-One
Working on a project for HC-One, I developed analytical and forecasting solutions to address
the challenge of receiving 2,500 - 3,000 weekly job applications for care workers and nurses.
The project targeted technical artefacts, including exploratory data analysis using R, interactive
visualisations and dashboards in PowerBI, clustering, and predictive modelling in Python, to indicate
the likely success of job applicants having.
View More
Technical data analysis of cloud-based visualization for rendering terapixal generated by public IaaS.
Performed data analysis using Python on computational resources to generate terapixel of 3D city
visualization and environmental data from Newcastle Urban Observatory on public IaaS cloud GPU nodes.
Conducted detailed research on interplay between GPU temperature and performance, and the effect of power
draw on render time. Provided insights for optimizing cloud-based visualization and improving energy efficiency.
View More
Big data analysis using Spark and Neo4j for New York Taxi Fleet Restructuring
Spark Framework based analysis on a Big data of 132M New York City (NYC) taxi records for a NYC-based taxi company.
The aim was to provide insights into restructuring its fleet allocation strategy outside the Manhattan borough.
The analysis successfully identified regions with less competition due to Covid-related travel pattern changes.
Additionaly, used Neo4j graph database for centrality analysis to identify top 3 regions with highest centrality score
within each community cluster for optimizing resource allocation.
View More
Music Recommendation System based on user Facial Emotional Expression
Developed a Facial Emotional Expression Recognition application using Keras
Convolutional Neural Network with 7-class (primary emotions) classification.
Achieved 82% recognition accuracy on testing data for 40 epochs, utilizing NVIDIA Jetson Nano for training.
Employed a range of technologies including Pandas, Numpy, Matplotlib, Keras, Tensorflow, OpenCV, and Flask.
View More
Sales Insights Dashboard using Power BI
Developed Sales Insight Dashboard for a computer hardware supplies business organization based in India.
The dashboard utilizes a MySQL database to extract and organize sales data, and Power BI to generate
comprehensive data visualizations. The dashboard provides the sales director with data-driven insights
to make informed decisions that can increase sales.
View More
Driver Drowsiness Detection System
A driver drowsiness detection system is designed to detect whether a person is feeling
sleepy or not while driving based on their behavior of yawning. The system uses transfer learning with a
pre-trained InceptionV3 model architecture for efficient pattern recognition and classification.
The live feed from the system's camera, which can be a dashboard camera for vehicles, is used to
determine a person's state of being. The model is trained for daylight, darkness, and multiple view
angles of the camera. The model has an testing accuracy of 90% and can be implemented in vehicles to reduce
road mishaps. The project utilizes technologies like Pandas, Numpy, Matplotlib, Keras, Tensorflow, and OpenCV.
View More
Helmet Detection for Motorcyclists in Traffic
Helmet Detection System that uses live feed from traffic cameras to detect whether motorcyclists are wearing helmets.
The system employs Object Detection YoloV5 model and a custom dataset on Deep Neural Network to train a CNN
architecture model for image classification. The model has the ability to detect multiple helmets with or
without helmets simultaneously, and has achieved a helmet recognition accuracy of 80% for 150 epochs.
The system can potentially be used to enhance road safety and prevent accidents caused by riders not wearing helmets.
View More
Garbage Detection on Roadside and Public Place Litter
Developed a Garbage Detection System utilizing Deep Neural Network with Convolutional
Neural Network (CNN) architecture for real-time detection of roadside garbage and public place litter.
he system was built using TensorFlow-Keras with multiple layers for training and testing the model for
image classification. Achieved a testing accuracy of 81% after 25 epochs for binary classification of
garbage and litter. The system can be used in real-time applications to maintain cleanliness and hygiene in society.
View More
Covid19 Infection Detection using Chest X-Ray Images
Covid Detection System is a Deep Learning-based model that can detect Covid-19 infection in
lungs using x-ray images. The model uses pre-trained un-toped VGG16 model architecture and
includes DNN and Convolutional layers for binary image classification. With the use of Pandas,
Numpy, Matplotlib, Sklearn, Keras, Tensorflow, and OpenCV technologies, the model achieved a
recognition accuracy of 87% for 20 epochs. The system can effectively detect Covid-19 infection in
patients and provide an early diagnosis for prompt medical intervention.
View More
Previous Work
SparrowAI
Fastbadge
Vigilview
FellowFarmer
AddNectar
Individual Undertaken Projects
Google Cloud Platform
Deployed a Multi-tier Web Application over GCP(Google Cloud Platform) using VPC Peering,
Kubernetes Cluster, Load Balancer and SQL Database service of Google Cloud.
Docker
Launch or Terminate/Remove DOCKER Container on Virtual Machine by Voice Control using Web API
Swift
Responsive Airline Ticket Booking Web-Application with provision to Book Airline Ticket.
This Web-Application provide Public view and Admin Management view.
Wanderlust
Created Interactive Wanderlust Travel Blog Application with provision to Read,
Comment, Share Blog and Create new Blog for Registered User.
Additional Experience
Software Carpentry Workshop Helper, Newcastle University
Assisting software engineering instructors by helping 30 researchers with computing skills required for
their research by explaining complex technical concepts, helping researchers with coding exercises and
programming concepts and syntax in Python, setup Git and explaining concepts and activities involved
in version controls, helping participants understand the basics of Linux commands and troubleshooting
any technical issues that arose in the software carpentry workshop.
