Ashirwad Sangwan

Data Scientist

• Propensity Model: Built and deployed a classification model for a client where the client wanted to know which segment of customers would opt for credit cards provided by one of the partner banks. Created and deployed the model in AWS Sagemaker.
• OTA to Direct Booking: Leading a team to solve a problem of converting customers who book their hotel stays through OTA platforms to booking platforms provided by our client.
• Next Stay Prediction: Created and deployed next stay prediction model for a hospitality client to find the right segment of the people to target for marketing. Used Mean Absolute Error as the model metric for optimization.
• Chatbot using QnA BERT: Created a chat bot for one of the top Hospitality business in the world that is being used by millions of customers around the globe. Used BERT question-answering pre-trained model and tweaked it for our use-case to make the chat bot experience realistic for the customers.
• Automation using NLG: Leading a team in developing a logic which will read data from an excel file and gives an output in the form of smart insights to make business decisions. Using transformers, we generate language from a processed input which comes from the raw excel files.

Senior Analyst

• Price Recommendation: Developed a price recommendation model for food items using SLSQP algorithm for a fast-food client. Used Price elasticity, item price and units sold to formulate the objective function for Revenue maximization. Planned back-testing to fine-tune the model for better recommendations.

Analyst

• AutoML: Created data science accelerators which are being used by the whole company to fasten the process of the model building for a business problem. Documented and Tested the code for our company product which has been in use by multiple clients.
• Predicting Store Location: Built a model to predict the optimal new locations for a Retail client that wanted to find the optimal places to open their stores to maximize the profits.Used linear regression to get revenues at new locations depending on demography and other variables.Then depending on the revenue predictions for each of new locations, optimal locations were suggested.
• Outlier Detection and Model Evaluation Tool: Built an Outlier detection module inside the python library for a Research and Development project to get more insights during exploratory data analysis. Also, created a model evaluation tool which generates insights about the model building processes like evaluation metrics, feature importance and model comparisons etc. for multiple model input.

Data Scientist

• Winning Algorithm and Customer Retention: Worked on designing a framework that helped us gain double customers in two months’ duration. Designed and implemented a retention cohort that takes up multiple factors into account to define the product growth metrics.
• Customer Segmentation: Used Clustering methods to segment the customers into multiple segments to help the marketing team target the right set of customers.

University of Delhi