Developing a Machine Learning model to optimize the process flow of fixing bugs in the SAP system. Different models from scikit-learn were used for learning the old customer issue/incidents. Different kind of vectorizers were used to vectorize the text from the reported bugs. Initial EDA is done and followed few steps, such as stop words removal, lemmatization, stemming & domain abbreviations

Roles and Responsibilities:

• Worked on text mining and information retrieval from the data provided by using BOW, TF-IDF, Word2Vec, TFIDF Weighted Word2Vec.
• Performed EDA and built machine learning models using RandomForest, XGBoost, SVM, Logistic Regression, k-NearestNeighbours, NaiveBayes, etc
• Understanding the performance of various algorithms used above.
• Optimizing and Improving the algorithms performance

Developing a text analytic model of customer messages raised for the Insurance product sales. Given the messages we have to predict the nature of the customer like message processing statistics, Behavior of the customer, whether the request turned into sales.., etc.

This product would provide the capability to guide the user to choose correct product for their requirement and guide them with correct configuration available in the sales using Artificial intelligence and machine learning. NLP and Text mining are used to guide the user with the correct product and also we use clustering and classification algorithm to find out the correct configuration of a particular product from the historic data. We also used Association rule machine to suggest related product to the customer.

Roles and Responsibilities

• Performed EDA on customer profile data and drawn insights on trends different kind of product preferences using different plotting techniques.
• Analyzed the customer behavioral patterns, time for which the product sales right from initial request to final purchase of the policy.
• Analyzed the reviews/feedback given by customer using various NLP techniques.
• Applied various Machine learning techniques like Linear Regression, SVM, Naïve Bayes, Clustering techniques.
• Applied association rule learning by categorizing customer data and tried to suggest similar products.

• Knowledge on withholding tax functionality concepts according to Countries legal requirement.
• Involved in message solving related to functional queries for countries like US 1099 reporting, SICORE reporting for Argentina, WHT Reporting - Spain, Italy, Belgium, and Deferred Tax for Thailand etc.
• Involved in solving customer messages for standard withholding tax reports such as RFIDYYWT, RFKQST00, RF0KQST5, J_1AF016 and RFWTCT10.
• Modifying the standard reports according to legal requirement for country specific functionalities and Releasing notes.
• Developed Argentina report for Exception file processing entirely using Agile Methodology.
• Involved in re-designing of withholding tax report for Italy.
• Involved in creation of new Utility report for Argentina RG2616 tax regime.

Good knowledge on Argentina tax

• Our team is responsible for maintaining and developing internal software’s that are required various operations which can reduce manual work, Fraud Detections such as Material gates, NOCL concrete supply, Bill tracking system, leave management system and contract labour management system etc.
• With the NOCL concrete supply we reduced the gaps between concrete supplier, concrete supply in-charge and contractor in the plant, where they can check each and every dispatch of concrete to contractor and also generated reports based on different selections.
• Received a letter of appreciation from Site in-charge for successful go-live of NOCL Concrete supply management system.

Building a simplified self driving car where to model we give sequence of images (video here we have 25 minute long) and output model gives is sequence of predicted steering wheel angle. It is a regression problem which we impose to Deep learning problem.

We are having bunch of images taken by Car front Camera @30FPS(30 photos Per Second) and corresponding angle to it ,we are having approximately 45k images , amongst 45K images we took 70% as train and 30% test , we will train the CNN network using Training set and then we will predict the angle of steering in the test data.

Building a machine learning model that recognize the human activities based on the 3-axial linear acceleration and 3-axial angular velocity readings of accelerometer and gyroscope, a activity can be any one of the WALKING, WALKING_UPSTAIRS WALKING_DOWNSTAIRS, SITTING, STANDING, LAYING

Amazon Fine Food Reviews is a classic Sentiment Analysis problem used to classify the polarity of the review given by the Amazon user. Performed data optimization tasks such as data mining, data cleaning, exploratory data analysis, review text preprocessing using pandas, sqlite3 and numpy. Utilization of NLP techniques such as BOW, TFIDF, Word2Vec to convert text data to vectors. Used effective algorithms such as K-NN and Naive Bayes to classify a given review. Applied cross validation to split the vector data into train and test data sets and also used cross validation to obtain the best value of K and Alpha on the metric of accuracy score. Finally tested the data with the model developed on the most recent reviews using both KNN and Naive Bayes model.

Our team is responsible for maintaining and developing internal softwares that are required various operations which can reduce manual work, Fraud Detections such as Material gates, NOCL concrete supply, Bill tracking system, leave management system and contract labour management system etc. With the NOCL concrete supply we reduced the gaps between concrete supplier, concrete supply in-charge and contractor in the plant, where they can check each and every dispatch of concrete to contractor and also generated reports based on different selections. Received a letter of appreciation from Site in-charge for successful go-live of NOCL Concrete supply management system.