I'm a Data Scientist living in the heart of Bangalore, India. I have over 3 years of experience working in data science. Currently, I work as a Data Scientist in Allstate India, improving products and services for our customers by using advanced analytics, creating and maintaining ML models, and onboarding compelling new data sets.
Previously, I was a Data Scientist at Wipro Holmes, helping clients use machine learning models to enhance customer experience and reduce cost
Competencies: Data Science, Machine Learning/Deep Learning, NLP, Computer vision, Python/Java/R, SQL, Pytorch/Tensorflow/Keras/scikit-learn, numpy/pandas/scipy/opencv
Rohit Raj
#1425, BTM 2nd stage
Bangalore 560076 India
(+91) 9738899609
devrohitraj@gmail.com
Bachelor of Information Technology • June 2015
Data Structure, Algorithm Designing, Object-Oriented Design, Java Programming, Database Systems, Networking, Distributed Computing, Artificial Intelligence.
•Co-Founded the Technical Club of the college that organized the 1st Technical Fest of the college. •Successfully assisted the College Placement Cell in attracting twice as many companies during hiring season than those participated the last year. •Served as a core member of the Planning Team at the college that organized its first ever Alumni Meet in over 10 years.
Class 12• March 2009
•12th from PCV Mahavidyalaya, BSE Board, Chhapra, Bihar with 66%.
Data Scientist - Artificial Intelligence Center of Excellence • Mar 2018 - Present
•Distracted driving model using mobile and telematics data to reduce risk of accident that can claim lives and dollars for insurers and insured. Deep learning computer vision model with 99% recall having 88% precision •Defect recommendation engine for product teams to predict potential defects that could occur in future projects resulting reduced hours of corrective work and reducing cost ~$50 million per year •A real-time recommendation engine for technical support team for query resolution resulting reduced average wait time from 12 min to 2 min. Hierarchical deep learning model using LSTM/GRU with ~99% accuracy •Helped team with various setups like Domino, UNIX dev server, AWS Sagemaker, Server DB, model deployment and so on
Data Scientist - Machine Learning • Nov 2015 - Mar 2018
•Implemented a Bot that creates a knowledge graph using spacy, genism and other NLP packages using Machine learning •Trained a ML model using SVM that classifies different element inside images with more than 75% accuracy on very limited data •Implemented AdaptiveUI Bot Agent using Python and MongoDB for Recording the test cases and process it for playback •Implemented Tracer and git-diff features for Regression Testing Bot using Java technology that reduced time by 30% and cost by 40%
Software Engineering Intern • Sept 2014 - Nov 2014
•Designed and implemented Employee Management Service, which managed the working of employees by providing reports between 2 days. •Implemented a centrally-controlled Administrator Module that performs all system administration tasks using Java Servlet Backend Technology with MySQL Database. •Utilized JSP as a view component of MVC architecture. •Tested the fully developed project EMS and discovered 7 bugs, including SQL injection problem for which received appreciation from the Project Manager and was asked to assist senior developers in debugging.
•Reviewed unstructured data to understand the patterns and natural categories that the data fits into and implemented model using PCA and KMeans. •Made predictions about the natural categories of multiple types in a dataset, then checked Against the result of the unsupervised analysis. •Implemented feature scaling and obtained 0.44 silhouette score to get efficient number of clusters. GitHub Repository of this project
•Analysed the student intervention dataset to model the factors that predict how likely a student is to pass their high school final exam and then by using that factors implemented the tuned model to predict success and failure early enough to stage effective interventions. •Built a Student Intervention System to suggest ways to increase school graduation Rate From 67% to 95% (Python) •Analysed 31 parameters to build a model to identify students who need help before they Drop out of school. •Utilized F1 score as the performance metric to compare different techniques- SVM(F1=0.78) Decision tree (F1=0.67) and KNN (F1=0.73). •Improved F1 score of the best model by 2% by model tuning(Grid-search). GitHub Repository of this project
•Builded an optimal model based on statistical analysis on various features of Boston housing dataset by using Machine Learning algorithm to predict selling price of homes in Boston, Massachusetts (Python). •Utilized Boston Housing dataset to build the optimal model to estimate the best selling price of Boston homes. •Exploratory data analysis to visualize the effect of different features on house price. GitHub Repository of this project
•Used Kaggle European soccer dataset to predict footballer overall rating by using physical attributes and game attributes like height, shooting power etc... • Used pandas get_dummies to convert categorical data to numerical data. •Used sklearn MinMaxScaler to scale to feature data. •Used Decision tree to train the features and get accuracy of 95% on separate testing set with little slow training speed (~6 sec). •Used SGDregressor to train/predict features with 85% accuracy but with incredible training speed(~0.8 sec). GitHub Repository of this project.
•Used Kaggle Breast Cancer Wisconsin (Diagnostic) Data Set to predict if cancer is malignant or benign. •Visualize the dataset for learning and complexity performance with Random Forest and used validation curve to find best cross_validattion score for parameter tuning. •Performed feature engineering on given dataset and get 93% f1_score and 92% cross_validation score with Random Forest on separate testing set. GitHub Repository of this project
If you have any questions about my work or just want to say hi, don't hesitate to contact me.