SHUBHAM MIGLANI Website LinkedIn GitHub
Email - adgqho@r.postjobfree.com Contact - 984-***-**** Address - Raleigh, NC 27606 Professional Summary
Result-oriented individual with strong analytical and programming skills. Experienced in utilizing statistical analysis, data modeling, and building machine learning pipelines to solve challenging problems and effectively communicate results. Education
North Carolina State University, Raleigh, North Carolina 4.00/4.00 Master’s in Electrical Engineering Jan 2019 – Dec 2020 Punjab Engineering College, Chandigarh, India 8.85/10 Bachelor’s in Electrical Engineering Aug 2012 – May 2016 Skills
Programming Python, R, SQL, C, C++, Matlab, Simulink
Frameworks & Libraries Scikit-learn, Pandas, TensorFlow, Keras, NumPy, Pyspark, XGBoost, nltk, spacy, OpenCV, ACL
Relevant Skills AWS–S3 & Machine Learning, Apache Spark, MongoDB, Google Data Studio, Tableau, Docker, Git Work Experience
MathWorks, Intern, Natick, MA May 2020 – Aug 2020
Developed workflow to switch between different domain libraries (OpenCV, Arm-Compute) during C++ code generation for image processing functions leading to 2 times performance improvement on arm-based processors. C++, Matlab WizeView, Research Assistant, Raleigh Jan 2020 – May 2020
Generated datasets and trained OCR model to convert drug label images to text & analyzed image processing techniques to improve performance. Python, AWS Rekognition, Tesseract-OCR
Trained and evaluated NER (Named Entity Recognition) models to identify drug names from the OCR text. Models trained: Memory tagger, Random forest, Conditional Random Fields, Sequence tagging (LSTMs).
Deployed model on iOS with a final accuracy of 76% from 50%, an average time of 1s from 2.24s after hyperparameter-tuning. Sabre Travel Technologies, Software & QA Engineer Intern, Bangalore, India Jan 2015 – Jul 2015
Built automation scripts for Data validation & performance testing for 55 Jasper Soft Reports. Java, MySQL, JIRA
Modified 80 scripts to include recovery scenarios for error handling & improving maintainability, increasing productivity by 60%.
Utilized dynamic SQL queries for automated testing of business rules for RM GUI. MySQL, QTP, VB Script Fiat Chrysler Automobiles India, Assistant Manager, Pune, India Jul 2016 – Aug 2018 Academic Experience
Graduate Research Assistant, ADAC Lab, NCSU, Raleigh May 2019 – Aug 2019
Built data infrastructure & developed visualization software for battery data for Smart Battery Gauge. Python, SQLite, Bokeh Graduate Teaching Assistant, Modern Control Systems, NCSU, Raleigh Jan 2020 – May 2020 Academic Projects
CNN for Leaf Wilting Detection
Developed CNN with transfer learning (72% accuracy), Improved accuracy to 77% with semi-supervised learning (unlabeled data)
Deployed model as REST API with flask, improved inference speed by 86% using tflite with quantization optimization Book Recommendation System
Implemented Popularity-based, TF-iDF, User & Item-based Collaborative filtering, MLP models for book recommendation system
Designed a multimodal (CNN+MLP) approach utilizing book covers with categorical data to improve performance by 1.4% Face Detection and Recognition
Face image classification with Gaussian, MOG, T-distribution, & Factor analyzer. Best model AUC score: 0.94
Implementation of Cascade of Haar feature classifiers with Adaboost ensemble learning for Face Detection (78.5% final accuracy)
Built a Face Recognition and Verification system utilizing pre-trained Inception v2 for encodings Customer churn prediction using Spark
Performed exploratory data analysis, feature engineering, and predictive modeling for churn prediction utilizing Apache spark
Models trained: Logistic regression, Decision tree, Random forest, and Gradient-boosted trees. Best model F1-score: 82.3% Reinforcement Learning - Optimal Control of Human-Robot Interaction system
Solved the LQR problem for unknown human-robot interaction using the actor-critic method for integral Reinforcement Learning
Conceptualized and implemented Neural Network for varying human parameters to get an optimal solution of the LQR problem Image to Image translation through Conditional-GANs
Investigated conditional adversarial networks as a general-purpose solution for image-to-image translation with different Generators
(ResNet, UNet), Discriminators, and loss functions.