Projects

Here are some of my notable projects.

Design Portfolio

Design Portfolio

FYP - (Final Year Project 2020-2021)

Sales and demand forecasting, customer conversion, and expected show rate projections are an integral part of a business, like TRA(Travel Resorts of America), that directly impacts revenue and day-to-day business operations. For forecasting of sales and projection of bookings, there are several factors that are involved and can be answered by applying different analysis techniques to the leads data. This project helped the company to estimate customer demand in advance and make pre-arrangements accordingly, and find out the factors that convert a booking into a tour, and then convert a potential customer into a permanent member while doing sales forecasting to increase revenue, improve customer experience, and reduce complaints.

In this project, we ran 360+ ML, DL, and rolling-average based experiments to find out best model for sales forecasting, and finally proposed an ML-LA(Machine Learning and Longitudinal Averaging) based ensemble which gave us 96% accuracy with the best AUC-ROC, and other metrics score among all other 360+ models.

Tech Stack: Pandas, Tensorflow, Keras, Python, Google Colab, Sklearn, Numpy, Docker, FastAPI, Streamlit, Elastic Search (Kibana), Matplotlib, SNS, etc.

ML-Powered Product Title Chained Classification, on Lazada's Dataset - (Information Retreivel/NLP Project 2020)

Combined the concepts of Machine Learning, DL(Recurrent Neural Networks architecture), and Information Retrieval techniques with complete data and natural language preprocessing pipeline to solve multi level-category classification(Lazada Dataset with 37000 title records) problem with 87% accuracy.

Tech Stack: Tensorflow, Keras, Python, Google Colab, sklearn, pandas.

To see the source code, Click here!

Image Classification using Convolutional Neural Networks (CNN) - (AI Project 2020)

Implemented complete pre-processing pipeline and trained model for Image classification on CIFAR-10 data set using Convolutional Neural Networks, and achieved 85%+ accuracy.

Tech Stack: Python, TensorFlow, Keras, Google Colab, matplotlib.

CV/resume Recommendation System - (SWE/DB/AI Project 2019)

Followed Agile Software Development Life Cycle(SDLC) approach to recommend CVs based on their score(TFIDF) when Recruiter posts a job with the number of required employees and job descriptions. The app works like an ATS which is used for recruitment purposes, treating documents as vector spaces.

Tech Stack: HTML, CSS, JavaScript, Flask, MongoDB, and ReactJS, JWTdecode, pdfminer.

To see the source code, Click here!

Comparison Between Multi-Threads & Multi-Processes - (Operating Systems Project 2019)

Compared 5 series to exhibit a contrast between multi-threads and multi-processes in Ubuntu (Linux Operating System).

Tech Stack Linux/Unix- Ubuntu, C, fork methods, VMware.