HARSHIL KETANKUMAR CHAMPANERIA

adercv@r.postjobfree.com +1-480-***-**** linkedin.com/in/harshil-k-champaneria SUMMARY

Masters student with professional experience, entrepreneurial spirit and passion for Statistical Data Analysis, Data Science and Operation Research.

EDUCATION

Arizona State University, AZ, USA [August 2018 - May 2020] M.S., Industrial Engineering, Specialization: Industrial Statistics GPA: 3.94/4.0 Pandit Deendayal Petroleum University (PDPU), Gandhinagar, India [August 2014 - May 2018] Bachelor of Technology, Mechanical Engineering GPA: 3.3/4.0 TECHNICAL SKILLS

Programming Languages: Python, Java, SQL, JavaScript, R, AMPL Statistical Tools: SAS, JMP, Minitab, MS Excel

Visualization Tools: Tableau, Power BI, R ggplot2, Amazon QuickSight Database: MySQL, SQL Server, PostgreSQL, MongoDB, Amazon DynamoDB, Microsoft Access Cloud Technologies: GCP, AWS Lambda, Glue, SageMaker, S3, Redshift, Athena, CloudWatch, Mturk Tools: Apache Kafka, Air

ow, Terraform, Visual Studio, Jupyter Notebook, Agile, git, Jira PROFESSIONAL EXPERIENCE

Graduate Research Assistant, Arizona State University [August 2019-May2020]

Developed the deep learning models on top of the FaceNet algorithm to identify and quantify the bias in face recognition problem because of the unbalanced proportion of the genders in the training set

Developed an emulator enforcing AI-Deferral Structures to advance appropriate trust in a Face Matching task and validate the level of di culty through AWS Mturk service

Data Science and Big Data Analysis Technology Intern [Summer 2019] Apollo Education Group, University of Phoenix Arizona, USA

Built Data Pipeline of Email Disposition data on AWS using Apache Air

ow for data pipelining, s3 for data lake, AWS Lambda for data processing (from txt to Avro), Glue for ETL operation using spark and Redshift for data warehousing

Implemented Campaign-Level and Student-Level aggregations on email disposition data using pyspark on AWS

Achieved the deployment of the data pipeline to AWS Production environment using Terraform by HashiCorp Graduate Services Assistant, Arizona State University [January 2019-May 2019]

Developed a Testing Methodology for evaluating the Algorithmic Fairness of smart biometric technologies

Designed the reproducible model by training the softmax loss function of Pre-trained Inception ResNet model on google cloud using VGG dataset and validated testing methodology for determining the fairness of smart biometric technologies INDUSTRY PROJECTS

Generalized Lot Sizing - Production-Warehousing Model [Fall 2019]

Generated a Google Maps API key to access and retrieve data from Google maps by Python code to produce the distance matrix for 25 nearby cities to the metropolitan area of Barcelona in Europe

Developed a mathematical mixed-integer programming model, and using AMPL/CPLEX made optimal decisions to locate multiple production units and warehouses within the set constraints to meet the quarterly demands of each city Building Classi er for a given data [Fall 2019]

Given data was split into training and testing and training data was used to train a model using algorithms and classi ers like Support Vector Machines, Neural Network, Random Forest and Ada Boosting classi er

The model was tested by calculating the generalization accuracy rate and a balanced error rate along with the confusion matrix from K-fold cross-validation and the best model was selected that gave a perfect t for the given data out of all candidate models

Deep Learning of Brain Images for Classifying High-Risk Alzheimers Disease Population [Spring 2019]

[Funding Agency: ASU and Banner’s Alzheimer’s Institute (BAI)]

Implemented transfer learning/ ne-tuning concept on Convolutional Neural Network (AlexNet, ResNet) using TensorFlow to rst classify functional MRI data of AD subjects from normal controls or cognitively unimpaired (CU) persons at preclinical stage and then distinguish high-risk CU from low-risk CU at a preclinical stage

Achieved 70% Classi cation Accuracy on National Alzheimers Coordinating Centre (NACC) MRI scans Prediction of Monthly Electricity Consumption and Monthly Maximum temperature [Fall 2018]

Analyzed the temperature time series data and built Holt-Winter Forecasting Model and ARIMA Seasonal Model in R software to forecast future seasonal temperature in the city

Developed Transfer Function Model to illustrate the relationship between Electricity Consumption and Maximum Tem- perature and achieved forecasting accuracy of 97.9% using Seasonal ARIMA model in temperature forecasting

Contact this candidate