Nov 2018 – Jul 2020 · 1 yr 8 mos · Bengaluru, Karnataka, India

• Lead machine learning products like city recommendation, personalized notification and dynamic discounting.
• Worked on deep learning architectures like neural collaborative filtering, skip residual networks for tabular data, and contextual multiarm bandits.

Jan 2018 – Nov 2018 · 10 mos · Bengaluru, Karnataka, India

Jan 2016 – Dec 2017 · 1 yr 11 mos · Bengaluru, Karnataka, India

• App Recommendation using Duration Logs (2016) – Python, Spark.
• Using Samsung Big Data Platform logs to show usage based recommendation is better than install based.
• Popularity based recommendation and collaborative filtering are used in both cases. Used pySpark for distributed computing to handle large data
• Approximate Nearest Neighbor Search and Approximate Similarity Join (2016-2017) – Python, Spark.
• Used installed apps to find most similar user in platform.
• This is done by using minHash and Locality sensitive Hashing.
• Implemented LSH-minHash in Spark before the official release of LSH in Spark ML.
• User Clustering and segmentation (2017) – Python, Spark-ML.
• Formed user clusters based on their interest profile and time-series usage of apps.
• Used k-Means and MeanShift clustering methods and evaluated using parameters like Dunn’s Index and entropy.
• Building User Profile Based on Smartphone App Usage
• Using the App and URL usage logs developed interest, demographic, habit and location profile for each user using various machine learning techniques.
• Used positive unlabeled classification techniques for user modelling in the absence of ground truth data.

Jul 2014 – Jan 2016 · 1 yr 6 mos · Bengaluru, Karnataka, India

• As a software engineer I worked for various projects and in all of them I was required to learn the technology first and then accomplish tasks given.
• Implemented a matrix vector multiplication and matrix-matrix multiplication in GPU, optimized it to work faster than existing external library of CUDA
• Vision Assisted Location Tracking using RF (2015) – R, OpenCV, C++.
• Worked on a proof of concept method to show the increase in performance of RF tracking system when assisted by vision. Using statistical analysis showed how vision assistance will improve accuracy over standalone system