Worked with fraud analytics team and built predictive models using machine learning
techniques to determine fraudulent trades in equities, forex trades and wealth
management related transactions.
Used ensemble of ML techniques like GBM, RF, NN, etc. to derive the best fit model.
Combination of AUC and balanced accuracy score was used to evaluate the models.

Worked on determining Model agents by regression, clustering and classification.
 Generated insights and results by segmenting agents into action groups using statistical
analysis and algorithms.
 Identified steps that each action group needs to take to approach Model Agent.
 Ranked agents in each cluster based on the most appropriate criteria to determine their
relative value in terms of business output.
 Provide the client with specific, actionable steps to share with the agents to drive
improved performance.

 Incident ticket analysis for various platforms for identification of major issues and quick
redressal
 Network analysis of ETL jobs to develop optimized alerting mechanism
 Identification of rows majorly accessed by analysts so as to purge unwanted data for
optimized query search and analysis

 Built algorithm for an online platform to convert input in the form of natural language
from user, generate appropriate SQL from it and return the desired result
 Used fuzzy string search and cosine distance to match the table and column names
 Developed Naïve Bayes classifier model to identify database based on user input
 Extracted table and column names mapped to appropriate SQL query based on database
identified from classifier to return output.

 Developed a model to identify a future complaint depending on past interactions and
transactions of the customer.
 Formulated a mechanism to identify the mood of the customer and act proactively.
 Reduce the inflow of complaints in order to achieve more customer satisfaction.

Worked on building a data-driven model to provide supply-demand forecasting for BC
fulfilment needs.
 Built R-shiny app for taking ad-hoc requests from the user as input and display the
scheduling period.
 Forecasted sales values for all accounts for the next one year and accordingly forecasted
the Benefit Counsellors required.


 Worked on text processing for finding the root causes of server logs from ‘Service now’.
 Automated the entire support system with automated ticket classifier using machine
learning which reduced the total effort by almost 70%.
 Built a forecasting model to predict the hourly generation of tickets.
 Developed algorithm to generate alerts for long running jobs based on threshold values
of durations