Hillsboro, OR 765-***-**** email@example.com LinkedIn GitHub Website
Purdue University, US GPA: 3.73/4
MS in Industrial Engineering
Coursework: Data Mining, Predictive Modeling, Deep Learning, Computer Vision, Linear Programming, Information Engineering, Applied Regression, Application of Statistics, Supply Chain Analytics
Anna University, India GPA: 9.17/10
BE in Mechanical Engineering
Packages & Tools: Pandas, OpenCV, Pytorch, Apache Spark, Open3D, PCL, Microsoft suite, Power BI.
Certifications: The Complete Python Masterclass, SQL for Data Science, Algorithms & Data structures.
Data Analyst – Purdue University May ‘19 – May ‘20
Devised an ETL pipeline with batch processing for clustered data with 4 million observations using python.
Forecasted the resource workload demand of data center using ARIMA, LSTM and achieved a 10% error reduction.
Designed an interactive dashboard to visually track CPU utilization with Power BI and React and rendered solutions to switch to other Time Series algorithms.
Prediction of case status for h1b
Created a web application using the Django framework and used probabilistic SVM model to predict the probability of
H1-B visa approval and reported 80% accuracy to its truth.
Performed PCA, geo-coded the data using address interpolation, and fulfilled data wrangling with Apache Spark.
Built an interactive visualization using R-Shiny and redeemed the benefit of Exploratory Data Analysis using Tableau.
Object detection for autonomous vehicles
Implemented Dark flow YOLO for real-time object detection and processed point cloud to reconstruct obstacles in 3D.
Carried-out Transfer Learning for training a synthetically generated data and obtained a mAP estimate of 47.4 at 0.03 sec/image and justified the network’s performance against Mask RCNN based on the achieved metrics.
Credit card fraud detection
Detected anomalies to crack unusual patterns using KNN, k-means, Isolation Forest, SVM, and Autoencoders.
Executed Random under-sampling and SMOTE to tackle high class imbalance (99:1) problem and classified a fraudulent pattern using Neural Net, Logistic Regression, and Random Forests based on F1 score.
Visualized the data with 50 predictors using the t-SNE algorithm and obtained an AUC score of 0.984 using XGBoost, bettering the existing model precision by 3%.
Pattern matching database
Determined the affine transformation for image alignment of different exposures and misalignment of fewer than 5 pixels using a Multi-scale image template matching & sobel edge detection algorithm.
Validated this approach against the feature-based & intensity-based alignment techniques such as FFT, ORB & ECC and achieved distinguishable accuracy (87%) in matching a pattern.
Constructed a database of template images in SQL with a custom rule of matching threshold being less than 0.7.
•Deployed custom built classifiers - naive bayes, logistic regression, support vector machine, and decision tree with a limited data from a speed dating platform.
•Predicted success probability of the date for a potential future date with an accuracy of 85%.
Prediction of electricity consumption in the United States
•Leveraged Local Government Data and crafted a predictive model for forecasting the total electricity using R.
•Developed a Deep Neural Network (Multi-layer FC net) that presented a significant improvement in energy usage prediction (7%) and depicted the best course of action based on selected regions of interest.