Apr 2021 – Feb 2023 · 1 yr 10 mos

• AdTech Ranking: Built a contextual MAB that optimised both ad revenue (via impressions) and CTR by improving the vertical rank of adTech cards. This project was able to improve ad revenue by 6% and CTR by 1% as compared to the control set.
• Flights Insights: Built an insights generation tool for all customers in the GCC region.
• ODIN platform for contextual MAB experiments: Built an MLOps platform for creating contextual multi-armed bandits experiments. This supported multiple reward optimisation and off policy evaluation metrics.

May 2020 – Apr 2021 · 11 mos

• Cabs demand forecasting: Built a tool for cab demand forecasting and route insights

Oct 2018 – May 2020 · 1 yr 7 mos

• Hotels Dynamic Pricing: Built a dynamic discounting engine for the hotels vertical. It provides discounts for over 50% of all transactions (both domestic and international hotels), while, maintaining a better conversion-rate at a higher net-margin than the rest of the platform. In terms of impact, this project saved over $200k and generated $56M of revenue in one quarter with a conversion-rate of 16% (in domestic hotels alone).

Jun 2016 – Oct 2018 · 2 yrs 4 mos · Bengaluru, Karnataka, India

• Built and maintained the predictive engine which forecasted operational, clinical and financial time-series, enabling hospitals to judge the effectiveness of their marketing campaigns, enforce budgetary control and gauge the severity of seasonal viral diseases.
• Employed hierarchical forecasting upon grouped time series using models like ARIMA, exponential smoothing, XGBOOST, and NN-AR with a consistent prediction accuracy of 82-92%.
• Worked with unstructured medical documents to categorise and extract information regarding demographics, personal history, social habits, vitals, laboratory results, medications, diagnosis and procedures.
• Performed automatic medical coding of medical documents, including that of diagnosis to globally recognised ICD-10 codes and procedures to CPT codes, using stacked LSTMs with reasonable accuracy. Thereby, reducing human effort and automating a niche area of healthcare, traditionally, performed by experts.
• Built a Captain's Scorecard app that was utilised by top-level management on an everyday basis to track the transactions, P/L margin, and growth of each vertical business function in the company. Created using RShiny, this app also featured monthly projections of each vertical.

May 2015 – May 2016 · 1 yr

• User testing is an UX research platform that crowdsources UX for its clients. The clients often like to gauge the initial reactions of users from certain demographics for a website or mobile application that they would have created. Active feedback, testing for usability and accessibility were part of the mandate of being a UX tester.

May 2015 – Jul 2015 · 2 mos · Greater Delhi Area

• - Performed several cryptanalytic attacks on PRESENT, an ultra-lightweight block cipher for mobile devices.

May 2014 – Jul 2014 · 2 mos · Bengaluru, Karnataka, India

• - Worked on an in-house project for the HR department to determine the right fit for a new project based on a persons' résumé. We created an application that would enable the employees to upload their résumés which we would sort on a basis of skill, experience, and relevance to the new project. We worked on a Spring and JDBC framework.

May 2013 – Jul 2013 · 2 mos · Greater Hyderabad Area

• - Created an interactive web application that would enable us to visualise the results of teams working on a project in Solar Energy. We employed a TomCat server with webpage development in HTML and CSS.