Data Analyst

SAI TEJA SAGI

405-***-**** addzbm@r.postjobfree.com https://www.linkedin.com/in/saisagi/ https://github.com/saitejasagi EDUCATION

M.S Business Analytics Oklahoma State University, CGPA: 4.0/4.0 July 2018 - May 2020

• Presented a research paper titled ‘The Advent of Renewable Energy’ in Midwest SAS Users Group Conference, Chicago

• Received a full Academic Scholarship from the SAS and Teradata conference committee B.S Mechanical Engineering Indian Institute of Technology, India, CGPA: 3.91/4.0 July 2011 - May 2015 SKILLS AND INTERESTS

Languages and Tools: Python, R, SAS, SQL, Alteryx, AWS, RedShift, Git, Cassandra, PostgreSQL, SAS Viya, SSIS, SPSS, Teradata Certifications: SAS Certified Predictive Modeler, SAS & OSU Certified Data Miner, SAS Certified Advanced GGGGGGG GGGGGeGGGG Programmer, Tableau Desktop Specialist ML/DL methods: Regression, Clustering, Decision Trees, SVM, KNN, Deep Learning, CNN, NLP, Computer Vision Packages: NumPy, SciPy, Scikit-Learn, TensorFlow, Keras, PySpark, NLTK, spaCy, Imblearn, Tsfresh WORK EXPERIENCE

Graduate Research Assistant Spears School of Business, Stillwater, OK Jan 2019 – May 2020

• NLP project: Analyzed the tweets and performed sentiment analysis on the tweets of #MeToo movement

• Computer Vision project: Created an application using VGGFace model to identify deepfake images with an accuracy of 88%

• Recommender Systems: Built a collaborative filtering-based engine on customer reviews to predict the ratings given by users to restaurants. Used Alternating Least Squares (ALS) technique to achieve an RMSE of 0.12

• Big Data project: Identified chemicals that patients are exposed to based on clinical symptoms at the University of Southern Carolina Big Data Case Competition

Data Science Intern Cabot Microelectronics, Aurora, IL May 2019 – Aug 2019

• Created automated tools in SAS Enterprise Guide to evaluate the process efficiency of the HDFS database

• Decreased the processing time by 53% by optimizing existing SAS and python queries

• Introduced autoencoders instead of principal components for reducing the dimensionality in the production data

• Performed data integrity check, and data quality investigation between data from oracle and cloud-based historian servers Production Data Analyst Reliance Industries Limited, India July 2015 – July 2018

• Created dynamic Tableau dashboards highlighting KPIs and prepared ad-hoc reports which catered to management team

• Built a classification model using SVM to predict if the actual field estimate amount (FE) will be higher or lower than the authorized field expenditure (AFE) amount with an accuracy over 70% beating the previous model

• Achieved record low inventory levels (down by 17%) through turnover rate studies and budget allocations

• Developed ETL mappings, sessions, and workflows to provide aggregate report tables using SSIS and Informatica PROJECTS

Proposing a new loss recovery method – GM Financial (SAS, R, Tableau)

• Analyzed the decile/outsourcing strategy for recovering loss amount from defaulters and identified unfair treatment in accounts with same recovery scores

• Proposed a more efficient method for recovering loss amount using unsupervised clustering techniques An alternative route to curbing Opioid Crisis – A predictive analytics approach for SAS Global Forum (SAS, Python, R, Tableau)

• Took a 3-step approach to analyze state, county and prescriber level factors leading to over-prescribing of opioids by prescribers

• Created an H2o AutoML program and obtained an accuracy of 96% Identifying members at risk for continued long-term use of Opioids – Humana Mays Analytics Healthcare Competition

• Created a model using Gradient Boosting to predict if members will continue opioid therapy 6 months after initial prescribing

Classifying European Money Denominations - A Convolutional Neural Networks based approach

• Obtained a validation accuracy of 98% in classifying the 7 denominations of the European dollar using the pre-trained CNN models VGG16, ImageNet, and MobileNet

Predicting credit card fraud – A Kaggle competition on identifying fraud in credit card transactions of European cardholders

• Achieved a recall of 98% and precision of 77% on data with severe class imbalance by using non-parametric generative techniques such as KNN and Naïve Bayes

Contact this candidate