Data Scientist at Myntra
Aug 2017 - Present
Bad RTO prediction Sep 2017 – Present Project descriptionUsed customer's browsing data and click stream data to identify bad returns. ➢ Implement neural net architecture using keras to predict RTO’s ➢ Was able to reduce bad RTO's by 25% Dynamic Pricing for Myntra products Changing the prices on demand to improve revenue
Analytics Consultant at Bridgei2i Analytics Solutions
Jul 2016 - Jul 2017
Predictive Maintenance for printers Aug 2016 – Sep 2016 Project descriptionPredicted failure events using sensor data to reduce printer down time and to improve maintainability. 1. Used sklearn GBT classifier to predict failure events. 2.Class imbalances were removed by optimizing oversampling and under sampling factors. 3. Model validation was performed by train test set of (70-30) followed by 10 fold cross validation. Predictive maintenance of wind turbines Dec 2016 – Present Project description1. Trained an ensemble of RNN and tree based models to predict failure instances 10 days in advance so that the issue could be contained before it occur. 2. Predict the probable cause of failure Neural Net Research Team Jan 2017 – Present Project descriptionResearch, benchmark and document different neural net techniques like Auto-encoders , RNN and CNN to expand team's knowledge, particularly in predictive maintenance and image processing. Currently working with different cnn architectures to create reusable assets for image recognition .
Business Analyst at Amadeus Labs
Nov 2014 - Jul 2016
Next Generation Travel Intelligence (Nexti) Dec 2015 – Jul 2016 - Worked with customer loyalty data to create metrics around customer retention and customer conversion - PNR to ticket conversion probability to better reward the travel agents - Using historical data to recommend the most profitable seat sharing arrangements Make My Trip Customer modelling challenge May 2016 – Jun 2016 Project description1. Cleaned textual data to remove spaces and other unnecessary characters. 2. Selected the most important features using Random Forest model based selection 3. Used Randomized Search CV to tune the hyper parameters for Random Forest. 4. Final submission ranked 3rd. Social Media Analytics Oct 2014 – Dec 2014 Project description1. Accessed twitter REST API through python to pull daily tweets of leading hospitality providers 2. Performed data analysis on tweets to identify which tweets get retweeted/favorited most. 3. Created a strategy to improve the popularity of the tweets. 4. Used NLP to tag tweets as positive, negative or neutral.
B Tech in ECE at Delhi College of Engineering
Jul 2010 - Jun 2014
Win rate predictor system for a large US-based GPO Aug 2014 – Oct 2014 Project descriptionUsed numpy,pandas and sklearn library of python to achieve the following:- 1. Implemented logistic regression model that identified characteristics of deals with potentially higher win rates. 2. Used behavior data to not only segment and prioritize new deals for business action but enabling ongoing monitoring of deals over their life cycle. Speaker Recognition System Mar 2013 – Apr 2013 Project description1. Implemented mel scale and derived MFCC from the speech signal using Matlab. 2. Implemented Vector Quantization techniques necessary for voice identification. 3. Used LBG algorithm to correctly identify a speaker. Human Detecting System Nov 2013 – Dec 2013 Project description1. Used Haar like features and integral image to parameterize an image. 2. Used AdaBoost classifiers to detect human face.