Sharath M.K

Data Scientist with strong math background and 2.10 years of industry experience in solving real world problems using state of the art techniques like predictive modelling, data processing, Machine Learning and Deep Learning. Core skills include extrapolating data and providing actionable insights aligned with business needs, predictive modeling and statistics

Service Request Duplicate Similarity Problem: 1) Business Problem: Find the Duplicate Service Requests for which solution was implemented and suggest the existing solution to future SRs with same context. 2) Tasks: Performed Extensive Data Analysis on large datasets and used advanced visualization tools like T-SNE. Identified and developed new features using advanced mathematical and programming skills for building predictive models. 3)Use-case: Reduced the Time taken, and Workload involved to resolve the SR’s. TEK Systems August 2018 – July 2019 ✓ Developed the end-to-end Frameworks through Enhancement requests from Customer using Agile SDK. Suggest the price of the items on Online Platforms (Supervised ML) 1) It is a Regression Problem. The goal is to suggest the price of the itemsfor online platforms (like Mercari, OLX) where sellers want to sell pre-owned products. 2) Performed Univariate Analysis, Used Sentiment score on Product Description, Historical price statistics as features. Used TF-IDF Vetorization and applied Log Transformation on Target Variable to reduce the effect of outliers. 3) Used the Ridge, SVR, Light GBM and Stacking Regressor. Finally Light GBM was used to build final model as RMSLE was lowest compared to other models. Store Sales Prediction (Time Series Regression) 1) It is Time Series based Regression Problem wherein we need to predict the daily sales of Store for next six weeks. Store Sales are influenced by many factors like weekends, festive seasons, promos, discounts etc., 2) Performed extensive EDA, Identifying/Converting the stationary of the Time series data using Statistical test, Removed Multicollinearity using techniques (like VIF), developed exhaustive feature set for predictive Models. The Final Model used was Random Forest Regressor. (Kaggle: 0.419) Customer Feedback Analysis: 1) Business Problem: To Retrieve and analyze the oracle customer’s quarterly fee backs and identify most occurring common problems faced by customers 2) Use-case: This significantly helped company to id

Completed B.E in Computer Science