PROJECTS
I've done so many academic and non-academic real-time projects during my academic career that industry experts are already working on it. To gain knowledge and develop our technical skills, I believe that working on projects and doing research work will help to sharpen our skills.
Projects
At Present GenDigital
Product Genie Data Ingestion and Data Reporting Stack
Successfully implemented telemetry ingestion and aggregation for the genie product,delivering actionable insights on its performance to product teams and executive leadership team.
Automate LifeLock reconciliation process and design lifelock reconciliation report
Successfully setup data ingestion on adf and automated lifelock reconciliation process and delivering insights to lifelock team and olp team.
ADF OPS Patrol
Real time Teams ADF job Alerting setup to LiveWire Team to handle failure of the job.
Design and develop the CPRA compliance ETL business logic
Successfully worked on CPRA(California Privacy Rights Act) project for data cleanup activity and delivered 158 M accounts export to OLP team.
Active customer history performance dashboard
Successfully built aggregation job for active customer data and delivering data insights for internal team analysis.
Handle Adhoc Data engineering lifelock data requests
Sucessfully handled adhoc lifelock data requests to support lifelock team and delivered data insights to lifelock team.
Estore Remarketing data ingestion
Successfully implemented data ingestion for estore remarketing data and delivering daily data export to marketing team.
Datasets: GCP migration from azure delta lake.
As a part of GCP migration Successfully migrated datasets from azure stack to GCP for CDO team.
GCP BigQuery datasets OPS Patrol
Successfully implemented OPS patrol to validation check with the source on Bigquery GCP data catalog for LiveWire team.
Academic Projects
MASTER THESIS: STOCK MARKET PREDICTION USING DEEP REINFORCEMENT LEARNING
One way for resolving such complicated decision problems is reinforcement learning. For stock price prediction and optimizing expected return, reinforcement learning could be a preferable alternative. Deep learning approaches can extract features from data with a large number of dimensions. However, it lacks the ability to make decisions. Deep Reinforcement Learning (DRL) combines the Deep Learning approach with Reinforcement Learning's decision-making capability. By studying the practicality of applying deep reinforcement learning to stock markets, we aim to contribute to the advancement of financial AI. Even in computer science research, combining reinforcement learning with deep neural networks is a cutting-edge concept.
RESEARCH CASE STUDY ON ACOUSTIC BAT CALL DETECTION AND NOISE REDUCTION
In this project, we concentrate on using a deep learning approach to detect bat sounds using a number of audio recordings as our data sources, and at the same time, handling the background noises that are present in these datasets using a noise-reduction technique. We used the concept of mel-frequency cepstrums (MFCs) for our feature extraction. However, in our case, we initially faced the issue of imbalanced data which could greatly affect the overall performance of our model. We tried to address this issue by applying four resampling methods on our training data. To train our model, we used a simple feed-forward neural network and then evaluated our results with each of these resampling methods
SEMINAR PAPER ON REGULARIZED REGRESSION TECHNIQUES
In this project, I explained and examined the performance of the three Regularized regression models. Shrinkage methods enhance the accuracy of prediction. The value of the coefficients is reduced to zero using these methods. Shrinkage solves some of the problems associated with OLS estimates, such as multicollinearity, and tends to minimize complex problems of the models while avoiding over-fitting. For different penalties, the methods minimize a penalized Residual Sum of Squares. In these methods, the amount of Shrinkage is controlled by a parameter that is usually selected using Cross-Validation. I used plots and tables to show the characteristics of Regularized methods. According to the Regularized regression results, we might be able to identify performance differences with only moderate power. I compared the predictive accuracies of different models.
ADVANCED -SQL FOR BUSINESS ANALYTICS
Preparing data for the executive board meeting and Building a data-driven growth story for potential investors
List of implemented tasks: Traffic Analysis & Optimization Website Measurement & Testing Channel Analysis & Optimization Product-Level Analysis User-Level Analysis.
CODE RECIPES OF MACHINE LEARNING AND DEEP LEARNING IN PYTHON AND R
In this section, I have prepared the code recipes of different machine learning and deep learning models in python and R programming languages.
INTERACTIVE DASHBOARD USING POWER BI FOR BUSINESS INTELLIGENCE
In this project, I have performed various analytical tasks to generate reports. In addition, find insights from the database and transform the data and build a data model for reports in Power BI. I also have calculated different measures using DAX features in Power BI. I also have hands-on experience in time intelligence for business analytics.
This study aimed to provide information about nearby parking spaces for the driver and to make a reservation minutes earlier using supported devices such as smartphones or tablets, it is based on Internet of things. The present study proposes and develops an effective cloud-based SPS solution based on the Internet of Things. Our system constructs each car park as an IoT network, the number of free slots in car park areas will be transferred to the data centre. costs are decided according to the timing for which user is booking the slot. The SPS is based on several innovative technologies and can automatically monitor and manage car parks.