• Suggesting unexplored brand categories to users based on their activity( product views, orders and wishlist), catering to their potential interests• Improved CTR by 225% in comparison to banner in a similar position ,20% better PLP CTR than platform average

Developed Web Quote Pricing Models for the LATAM market (Chile & Mexico), currently live for 80% & 20% traffic respectivelyDesigned and implemented a cutting‑edge ranking algorithm to accurately determine the relative rankings of various trims for models of carsPioneered an innovative optimization algorithm that transforms the pricing strategy for second‑hand carsReduced the average error in sell price prediction to less than 3%Increased the per car margin by more than 30% in Chile Market

- Smart Recruiter (NLP)- Worked on Vision‑based Sectioning Algorithm to map different sections of resume based on coordinates of bounding boxes- Worked on NER model to extract different entity like Name, College, Degree, Past company, skills from the resume- Worked on Statistical method to extract sections from resume- Price Optimization- Worked on optimization problem to find the optimal price of a product based on existing user’s purchase behaviour• Clustering of user’s and used entropy‑based method to find the price that minimizes the entropy across the clusters• Profiled the users based on their purchasing behaviour using K‑means clustering algorithm and analysed the persona‑wise customers- Media forecasting & analysis- Extracted objects & text from Creatives using Google Vision APIs, performed Hypothesis testing & analysed the impact of different objects (person, animal, kid), word presence on the CTR performance of Creatives• Used Tensorflow pre‑trained mobilenetv2 model & OpenCV to map different video (child) to it’s corresponding parent video based on a similarity score

OBJECTIVE- Developed Package for Data Pre-Processing of Time Series Data which includes Data Exploration, Missing Value Treatment, Outliers Detection feature Engineering (Feature Creation and Feature Selection) consisting of multiple options for each kind.- Developed Package for Diagnostic &Prognostics for isolating Engine Failures, Fault Mode, Health Index, Predicting Remaining Useful Life in Python.ANALYSIS- Used the Developed package for doing Analysis on NASA Engine Data set and A-320 neo, able to produce a remarkable result in less amount of time and code. Used Relevence Vector Machine to to denoise the Health Index pattern with respect to time.- Used LSTM and RNN to predict Remaining Useful life of NASA Engine data set, got a mean absolute error of 10 useful life on test data .

OBJECTIVE- Used real-world geospatial data (including latitude, longitude, day and time) recorded using GPS devices to extract meaningful locations, and modelthem in such a way that we may predict where a user will be at a given time and day of the week .- Time series analysis of spatio-temporal data using Hidden Markov Model and Predicted future location of usersCLUSTERING- Analysed the GPS data of users using DBSCAN algorithm using Haversine as distance Metrics to calculate distance between two place.- Plotted the Cluster on Map and found place of Interest with respect to time and day and found interesting pattern that was repeating on only duringparticular time and dateANALYSIS- Divided the each day into 8 regular timestam .Used Bayes theorem to find Place of Interest of particular user during any day of the week at particulartime (Sunday at 7 pm).- Implemented Hidden Markov Model to predict future location of user given that the user sequence of travel