Worked on denoising & HDR+ IQ tuning for Video-Boost on Pixel 8 ProWorked on Super Res Zoom Video & 8K Video on Pixel 9 Pro & Pixel 9 Pro XL

Teaching Assistant for CS 4644/7643: Deep Learning taught by Professor Danfei Xu

Worked on the design and development of machine learning models for radar point cloud segmentation

Implemented a proof-of-concept salient object segmenter using superpixels to generate preview thumbnails for Samsung's smartwatches.Cleared the Professional level of Software Competency test conducted globally by Samsung Research.Implemented a human-segmentation DNN commercialized on the Galaxy S10, Galaxy Fold & Galaxy Note 10 devices for effect preview in the 'Live Focus' mode on the front camera.Trained the preview & capture segmentation DNNs for the Front-camera 'Live Focus' mode of the Galaxy S20 line of smartphones. The combined efforts of the vision research team at SRI-B resulted in the best DxOMark bokeh score of 70 points for a single-front camera system and was above a few phones using stereo/ToF depth-sensors with the front camera. The segmentation DNNs' accuracy was improved further for the Note 20 line of smartphones, and was eventually propagated to many of Samsung's F, A, M & S series smartphones due to its great segmentation quality and good inference time.Implemented the initial dataset and training pipeline for multi-class instance segmentation for the Galaxy S21's front camera Portrait mode. The commercial implementation of the S21's Portrait mode was called 'the most improved, most accurate portrait mode' by Marques Brownlee (MKBHD).Implemented & commercialized a temporally stable segmentation DNN for the 'Video Call Effects' feature on the Galaxy S21 line. A lighter version of this DNN was commercialized on the Galaxy S20, Note 20, Z Fold 2 and Tab S7 devices. I was given an annual award for handling this feature's AI components (mostly) by my own.Implemented a real-time DNN for single-person body-keypoint estimation with a focus on fitness centric applications.

Successfully integrated Apache Spark with Apache Lens on a YARN cluster, looking into various improvements offered by Spark as compared to previously used Hive (with MR as the execution engine) including Dynamic Allocation.Lens with Hive on Spark led to successful & faster execution of Lens queries on a YARN cluster. Also observed that general Hive-SQL queries ran faster on the same cluster.Committed a Weighted Driver Selector to Apache Lens for the purpose of distributing jobs in a fixed ratio between multiple drivers.

Created marketing campaigns and ran them to promote various AIESEC events, while communicating AIESEC to different stakeholders Ran integrated online and offline campaign to promote the 1st ever YouthSpeak Forum in IIT Kharagpur.Created the 1st event website of AIESEC IIT Kharagpur, www.ysfiitkgp.inLed & trained a team of 6 members in key areas to achieve targets in a cohesive & engaging mannerIncreased FB page engagement by 40% with integrated content that suggests and drives action while improving AIESEC's on campus brand image.Awarded the Best Department for Initiatives & Operations for the year 2015-16

Manufactured the team's first ever PCB assembly, a PDM board to regulate power delivery and implement electronic actuation of components.Implemented an improved wiring protocol for our car with an updated wiring harness.

Managed the Facebook page of Intugine Technologies.