Ajay Bhattacherjee

About Me

My name is Ajay Bhattacharjee. I am a Data Science and Machine Learning enthusiast. Currently I am working as a Data Analyst in HSBC Global Analytic Center, Bangalore, India, with almost 3 years of industrial experience in Machine Learning and Data Science & Analytics. I also hold 1.2 years of research experience of using Deep Learning in Computer Vision and Natural Language Processing in Computer Vision Lab of University of Hyderabad. I am a post graduate in M.Tech Artificial Intelligence from University of Hyderabad, India.
I have actively worked on Predictive Analytics using cutting edge Machine Learning techniques and algorithms (e.g Random Forest, Decision Tree, KNN, SVM, XGBOOST, Naive Bayes) to provide business solutions and data insights. I have levereged various Deep Learning techniques (CNN, LSTM, MLP, Autoencoders) to solve problems of Optical Character Recognition, Neural Caption Generation, Chatbot. Apart from these, I have done variuos self-projects on Supervised and Un-Supervised Learning.

Professional Experience

DATA ANALYST \| HSBC 2018 - 2021	LOSS PREDICTION MODELLING Currently working on developing Loss Given Default Model for predicting loss for home loans portfolio of Australia. various Regression Techniques (Regression Tree, Random Forest, Linear Regression, Polynomial Regression) are been used to study the performance. Tools used : Jupyter Notebook, Python, Numpy, Pandas, SKlearn CUSTOMER PAYMENT MISS DATE PREDICTION This project aims at reducing outstanding receivables through improved collections strategies. A huge time is spent in understanding the portfolio and framing the assignment into Machine Learning problem. Various Machine Learning algorithms (Logistic regression, SVM, XGBoost,Naive Bayes, Decision Tree) are used. Tools used : Jupyter Notebook, Python, Numpy, Pandas, SKlearn PROBABILITY OF DEFAULT PREDICTION Worked on the project to develop a model which ensures acquired customers are majorly non-defaulting. At acquisition (booking) level it eliminates the bad base of customers, reduces the net loss of the portfolio.To develop this model Machine Learning algorithms (Random forest, Logistic regression, Naive Bayes, SVM, Decision Tree) are used. Tools used : Jupyter Notebook, Python, Numpy, Pandas, SKlearn CHATBOT DEVELOPMENT Developed contextual chatbots for information retrieval and data analytics, uses open-source Natural Language Processing (NLP) library - RASA for intent classification, entity extraction and dialogue flow control. Tools used : RASA, Spacy, Python, Flask, HTML, Javascript DATA DEVELOPMENT Worked on integration of data to create standardized datasets from raw data-mart. This process has involved complex SQL queries to fetch and summarize data from data-marts into standardize format. Tools used : SAS, SQL
RESEARCHER \| COMPUTER VISION LAB (UoH) 2017 - 2018	OPTICAL CHARACTER RECOGNITION Implemented OCR system using Deep Neural Network - Convolutional Neural Networks(CNN). Extended the OCR with Transfer Learning by using pretrained Autoencoders followed by Multilayer Perceptron(MLP). Transfer Learning was found to be giving better result of 99.5% Tools used : Python, Keras, Numpy, Pandas, Spyder IDE NEURAL IMAGE CAPTION GENERATION Worked on automatic image annotation (image-to-text). Deep CNN feature extractor is used as an encoder for images and Deep LSTM Language model is used as a decoder for text caption. Model having Encoder-Decoder Architecture is trained on Flickr8K Dataset. The final model gave a BLEU scores of 40.4. Tools used : Python, Keras, Numpy, Pandas, Spyder IDE

DATA ANALYST | HSBC
2018 - 2021

LOSS PREDICTION MODELLING
Currently working on developing Loss Given Default Model for predicting loss for home loans portfolio of Australia. various Regression Techniques (Regression Tree, Random Forest, Linear Regression, Polynomial Regression) are been used to study the performance.
Tools used : Jupyter Notebook, Python, Numpy, Pandas, SKlearn

CUSTOMER PAYMENT MISS DATE PREDICTION
This project aims at reducing outstanding receivables through improved collections strategies. A huge time is spent in understanding the portfolio and framing the assignment into Machine Learning problem. Various Machine Learning algorithms (Logistic regression, SVM, XGBoost,Naive Bayes, Decision Tree) are used.
Tools used : Jupyter Notebook, Python, Numpy, Pandas, SKlearn

PROBABILITY OF DEFAULT PREDICTION
Worked on the project to develop a model which ensures acquired customers are majorly non-defaulting. At acquisition (booking) level it eliminates the bad base of customers, reduces the net loss of the portfolio.To develop this model Machine Learning algorithms (Random forest, Logistic regression, Naive Bayes, SVM, Decision Tree) are used.
Tools used : Jupyter Notebook, Python, Numpy, Pandas, SKlearn

CHATBOT DEVELOPMENT
Developed contextual chatbots for information retrieval and data analytics, uses open-source Natural Language Processing (NLP) library - RASA for intent classification, entity extraction and dialogue flow control.
Tools used : RASA, Spacy, Python, Flask, HTML, Javascript

DATA DEVELOPMENT
Worked on integration of data to create standardized datasets from raw data-mart. This process has involved complex SQL queries to fetch and summarize data from data-marts into standardize format.
Tools used : SAS, SQL

RESEARCHER | COMPUTER VISION LAB (UoH)
2017 - 2018

OPTICAL CHARACTER RECOGNITION
Implemented OCR system using Deep Neural Network - Convolutional Neural Networks(CNN). Extended the OCR with Transfer Learning by using pretrained Autoencoders followed by Multilayer Perceptron(MLP). Transfer Learning was found to be giving better result of 99.5%
Tools used : Python, Keras, Numpy, Pandas, Spyder IDE

NEURAL IMAGE CAPTION GENERATION
Worked on automatic image annotation (image-to-text). Deep CNN feature extractor is used as an encoder for images and Deep LSTM Language model is used as a decoder for text caption. Model having Encoder-Decoder Architecture is trained on Flickr8K Dataset. The final model gave a BLEU scores of 40.4.
Tools used : Python, Keras, Numpy, Pandas, Spyder IDE

Qualification

EDUCATION	SESSION	CGPA / Percentage
M.Tech in Artificial Intelligence	2016-2018	8.0
Masters of Computer Application	2013-2016	8.22
Bachelor of Computer Application	2010-2013	80%

Projects

To enhance my knowledge, I keep on playing with the datasets available in online repositories. Below are some of the self-projects I have done. I have worked on both - Machine Learning and Deep Learning Assignments (both - supervised, Un-Supervised problems).
In the project grid below - "Self Driving Car", "Object Detection - Food Item" are Deep Learning projects. Rest all are Machine Learning projects. And "Netflix Movie Recommendation System", "NewYork Taxi Prediction", "Apparel Recommendation" are Unupervised Learning projects rest all are Supervised Learning Problems.