Sakshi Suman

I'm pursuing a master's degree in Applied Mathematics at Northeastern University - College of Science with a concentration in Machine Learning and Statistics. My primary strength is Data Structure & Algorithm Implementation. Besides this, I hold a strong background in Linear Algebra, Calculus, Probability & Statistics. I'm looking for a long term career in Machine Learning.

Transfer Learning with MobileNetV2

Used pre-trained weights of MobileNetV2 Convolutional Neural Netowrk on ImageNet dataset. Modified the network architecture by deleting the top layer and adding a new classification layer. Performed training only on the new layer in order to create a binary Alpaca classifier to increase accuracy from 0 % to 99 %.

Matrix Factorization for User Rating Predictions

Derived update rules and implemented Weighted Alternating Least Squares for predicting missing user ratings of MovieLens data. Evaluated the algorithm using MSE and found that it is 62 % better than baseline model.

Data Modeling using Markov Chain

Performed Time Series Analysis of average runs of opening batters in baseball from 1871 - 2015 with a Markov Chain. Calculated autocorrelation between original time series and a simulated time series. Performed GoF test at 5 % significance level to determine valid states of Markov Chain in a two-step transition matrix.

Customer Experience & Data Analytics Project

Proposed and developed a Sentiment Analysis model to predict customer satisfaction on chats and emails using Logistic Regression and Naive Bayes models in Python and SQL.

Predator-Prey Mathematical Modeling

Modeled Predator (Bald Eagle) - Prey (Rodents) population growth using Lotka-Volterra equations modified with weak Allee effect and pesticide constant. Simulated population plots with/independent of time and improved the existing model accuracy to 94 %. Also calculated lethal limit for rodenticide usage.

Northeastern NEWS Updates

Developed a Google Chrome extension to get instant notification updates from News @ Northeastern portal using JavaScript, AJAX, HTML, and CSS.

Pelatro Solution Pvt. Limited

Software Engineer - Machine Learning, Jun 2019 - August 2021. • Implemented K-Means algorithm to predict the Next Best Action for customers. Achieved accuracy of 61 %. • Developed an interactive web application to analyse and report statistics for a Machine Learning pipeline. • Predicted the Customer Lifetime Value using a Markov Chain and achieved an accuracy of 76 %. • Optimized duplicate row detection algorithm using probabilistic approach; reduced time complexity from O(n^2) to O(n). • Containerized and deployed end-to-end applications on production servers using Docker.

Walkter Beacon Lab

Data Science Intern, Jan 2019 - May 2019. • Built a CountVectorizer NLP model for comparing a user resume with job descriptions. Automated resume matching process and decreased the time spent by recruiting team by approximately 80 %. • Designed an efficient user visit logging system to calculate the user retention rate and automated email system for an ATS. • Adapted Tesseract OCR's code, to increase accuracy in text-recognition for screen fonts from 50 % to 95 %.

REVA University

Teaching Assistant, Jan 2018 - Dec 2018. • Courses: Core Java, Object Oriented Programming, Mathematical Foundations of Computer Science I & II. • Promoted to Head TA in Fall 2018; led weekly meetings and supervised four other TAs.

Languages

Python, R, Java, SQL, MATLAB, HTML, CSS, JavaScript/TypeScript

Machine Learning

Regression, Classification, Clustering, Dimensionality Reduction, Decision Trees, Random Forests, Bagging, Boosting, Neural Networks, Feature Engineering, Principal Component Analysis

Frameworks

tenosrflow, PyTorch, Hadoop, Apache Spark, Flask, NumPy, pandas, Matplotlib, scikit-learn, SymPy, Jupyter

Additional

Git, Jenkins, JIRA, Docker, Excel, IntelliJ IDEA, PyCharm, VSCode