Sayesha Aravapalli
I have always been intrigued by the story hidden behind numbers and data. After graduation from the Indian Institute of Technology, I worked as an analyst for Aspect Ratio, Merck's centre for Analytics. In the role, I handled business strategy development for one of the biggest players in global pharmaceutical industry, Merck. This experience made me aware of the enormous financial impact of data driven analytics, and sparked an interest to master my skills in the field.
I am currently pursuing my Masters of Science in Business Analytics at McCombs Business School, UT Austin. The diverse course structure, along with focus on hands-on projects with global firms has exposed me to the immense opportunities that lie within data analytics.
In my personal time, I like exploring new places and reading. Feel free to connect on any of the platforms linked below.
Education
Red McCombs School of Business, The University of Texas at Austin
Indian Institute of Technology, Bombay
CFA level 2 passed (Chartered Financial Analyst - USA)
Experience
Visa
Fraud patterns keep changing frequenlty and it is important to capture current fraud patterns using most recent transaction data.Traditional supervised models don't perform very well. The aim of the project is to find semi-supervised/unsupervised methods to detect fraud patterns. In particular, we will be using DBSCAN and graph based methods to detect unusual patterns and alert stakeholders.
Aspect Ratio, Merck's Centre for Analytics
Working as an analyst for one of the biggest players in the pharmecutical industry, I had the opportunity to tell stories with data about the KPIs of drugs.I assessed impact of market events such as entry of competitor's drug and loss of exclusivity of Merck's drug on sales and revenue in the anti-biotic drugs space based on therapy area knowledge and treatment protocol. Collaborated with LATAM, APAC and Gulf marketing teams to design and create market opportunity analyser for Bridion(anti-anesthetic drug) to increase focus on creating more product value before its loss of exclusivity
DBS Bank
Stress testing is an important aspect for financial institutions especially banks. It measures the health of the institution and the ability to withstand major financial crisis. As part of the team responsible to deliver an integrated stress testing framework to client, I automated the operational risk module. I also identified opportunity areas for cross selling products by evaluating cross sell index of product exposure for both retail and corporate clients.
Skills
- Python
- R
- MySQL
- Hadoop
- Mapreduce
- Git
- iPython/Jupyter
- Google Cloud
- Google Vision
- Google Analytics
- Tableau
- Linear Regression
- Logistic Regression
- Naive Bayes
- Principal component Analysis
- Linear Discriminant analysis
- Random Forest
- Boosting
- K-means clustering
- DBSCAN
Projects
Google Analytics
A consulting project for a solar energy based start-up to understand potential customers based on customer sign-ups. The goal of this project was help the management understand features influencing customer sign-ups - Recommended marketing channels and optimization of mobile website. Also advised a third party vendor for easier access of data for future use.
Instagram User Engagement
Considering that in 2014 social users posted an average of 1.8 billion photos to the internet daily and the growth of image-centric platforms like Instagram, Snapchat and Pinterest makes analyzing visual data also very important to understand your audience.We tried to answer What should National Geographic do to increase engagement on its Instagram page?. Image analytics using LDA topic modeling on scraped data from the Instagram Natgeo handle, gave some insights into types of posts that are least and most engaging.
Twitter CEO Engagement - NLP
Twitter, with increasing followers, can be a social platform for CEOs to connect directly with customers. Analysed tweeting styles of CEOs from different sectors to ascertain the characteristics that make them influential. Stock price analysis (regression) showed influential CEOs can actually impact stock prices through tweets. Influential CEOs can leverage Twitter to turn negative sentiments of the crowd to positive ones and affect the market.
Recommendation Engine
Today in the world of personalized marketing, it is important to understand each customer and target them accordingly. The task of analyzing customer behaviour from humongous data is a challenging one and this project attempts to recommend books to users based on their given ratings of other books. Collaborative filtering methods on books clustered using K-means improved accuracy and a reccomendation of mixture of books from different clusters can help solve the long tail problem and increase customer satisfaction and experience.
Car Brand Association Analysis
Marketing is all about want you want your customers to associte your brand with.User generated content can give insights on consumers' perceptions. Car reviews data scraped from Edmunds forum helped understand most discussed car brands and most discussed attributes of cars using NLP. Evaluated associations between brands using lift analysis and visualised it using Multidimesional (MDS) plot.
US Traffic Fatalities Analysis
Fatality Analysis Reporting System (FARS) was created in the United States by the National Highway Traffic Safety Administration (NHTSA) to provide an overall measure of highway safety. It contains data on a census of fatal traffic crashes within the 50 States, the District of Columbia, and Puerto Rico. To be included in FARS, a crash must result in the death of a person (occupant of a vehicle or a non-occupant) within 30 days of the crash. Surprisingly, 47% of the fatal crashes happened during the day and just 27% of the accidents where due to drink drivers.
Austin Crimes
To understand how safe is the city of Austin, we set out to analyse Austin crime data and dig deeper into why crimes in Austin increased drastically in the year 2018. The data showed that non-violent crimes have been increasing but violent crimes were decreasing. The average processing time for violent crimes is 44 days where as for non-violetn crimes is 30 days.
Energy Consumption Prediction
Recent global phenomena related to climate change, call for a new strategy to collectively move towards sustainable energy and reduce our carbon footprint. In one such effort, building owners are incentivised to improve building efficiencies to reduce operating costs and emissions. In this project, building energy consumption is predicted which can then be compared to the actual running costs and decide upon the incentives. Operating costs from over 1000 building over a 3 year timeframe is used to build the prediction model.