Badges
Certifications
tulluri1 has not earned any certificates yet.
Work Experience
Software Engineer
invesco• June 2014 - July 2017
1.) Transition Intelligent Agent - Developed algorithm which analyzes incoming incidents and predicts support group based on the description in order to fasten routing process, developed by nltk, naive Bayes classifier with 89 percent accuracy which led to gain of 4800 hours/ annum of productivity. 2.) Improved customer experience based on subjective opinions of customers on enterprise product by accessing twitter API and developed using nltk, logistic regression, multinomial NB with L2 regularization resulted in 74% accuracy with 81 percent positive polarity. 3.) Improved response and resolution time KPIs from 7 days to 3 days by developing performance predictions and benchmarking model which compares with industry’s benchmark data set to drive critical business outcomes and to perceive our enterprise performance stacks up against with other peer’s outcomes from 5000k+ user records and 6+ billion transactions. 4.)Developed tracking algorithm which in turn saves 80 percent of the time on inventory with 91 percent of accuracy for predictive maintenance for assigned asset tags developed by nltk, logistic regression, SVM and random forest models which is rolled out to 4+ million asset tags. 5.) Automated skill inventory which relates internal job roles and key skills which needed for hiring managers by categorization of skills based on the education, technical experience, developed by regression, k-means clustering and association rule analysis which increased the productivity of 520 hours/annum in search of skilled employees. 6.) Decreased manager’s creation and approval time of timecards of employees by the factor of 30x (30 minutes to few seconds) whose status is out of the office with more than a week by analyzing the incoming data from people soft database which improves the productivity of a manager or a delegated approver by 60 hours/month
Education
University of North Texas
Data Science, MS• August 2017 - May 2019
1.)Microsoft Malware Prediction - Implemented an algorithm to predict probabilities of getting infected by various malware based on properties of machines by developing lgbm, sequential neural network models, by opting one-hot, frequency and boolean encodings with AUC score of 0.703. 2.)Crime Detection Algorithm-Formulated algorithm which reduces the crime rate about 20 percent developed by logistic regression, random forest (Ensemble) models for predicting the type of crimes and zip codes with the higher crime rate in order to increase patrolling. 3.)Mercari Price Prediction-Improved price recommendation algorithm based on the product category, descriptions, brand and condition developed by nltk, tf-IDF ridge and lasso regressions, and light gbm with rmlse of 0.74. 4.)Toxic Comment Classification-Automated classification of toxic comments in order to maintain fruitful platform conversations, developed by a recurrent neural network (lstm,gru), and CNN (fast text) with AUC of 0.992
Gandhi Institute of Technology and Management
Information Technology, BS• May 2011 - May 2015