Experience

SDE-2 Amazon, NY | Present

Designed and implemented an internal micro-services utilizing API Gateway, Step functions, Lambda functions, VPC, and EC2 API calls to establish cross-fabric connections for over 80k services.
Leveraged AWS Athena, Glue and S3 to merge large databases of 1M entries to extract detailed breakdown of $20M yearly cost by individual customers, ensuring accurate and comprehensive financial analysis.

Lead ML Engineer Samsung R&D |

Involved in the development and training of Human-Portrait model capable of running full HD image on Samsung Galaxy mobile at 60fps.
IAmAlpha Model developed for Portrait effect ended up commercialized in multiple Samsung’s Flagship mobile and is one of the top alpha matting model on alphamatting.com

Senior Software Engineer Samsung R&D |

Worked on calculating motion vectors among multiple frames using Block Search algorithm.
Enhanced finer details and reduced noise in image by utilizing Gaussian and Laplacian Pyramid blend.

Senior Software Engineer Samsung R&D |

Developed and optimized feature tracking algorithms for multiple frames, achieving real-time performance with frame blending under 300ms on the Snapdragon 835 platform.
Engineered and retargeted an imaging solution for heterogeneous platforms, ensuring it met Key Performance Indicators (KPIs) for mobile devices; successfully commercialized in Samsung Galaxy products for enhanced digital zoom and low light shots.

Software Engineer Samsung R&D |

Developed and integrated vectorized code for Noise Reduction algorithms with camera framework. Collaborated with various overseas teams to maintain image quality and improve performance by extensive NEON optimization.

ML Intern Samsung R&D |

This paper extensively investigated the impact of various dataset pre-processing techniques on the performance of deep learning classification models.
Implemented various pre-processing strategies and conducted thorough empirical evaluations, providing valuable insights to optimize workflows and enhance the performance of deep learning classifiers.

Projects

Task Orchestrator for Dynamic Edge Computing

Purdue University

Designed an algorithm for offloading tasks to Dynamic Edge Platform for reduced service time. Experimented with various clustering algorithms to group similar devices, improving scheduling efficiency and reducing end-to-end latency by 30%.

Dynamic Shape based Radio Tomographic Tracking

IIT Kanpur

Combination of Levenberg-Marquardt and RTI used to interpolate a spline over an object surrounded by wireless radios. Later, applied Extended Kalman Filter to track the object.

Graduate Teaching Assistant | Advanced C programming

Automated grading for a class of 400 for C-programming assignments and provided timely and effective feedback to the undergrad students.

Samsung Spot Award | Samsung R&D, Bangalore

Showcased Slow Motion and Night Shot projects in an annual tech fest at Samsung and awarded for Innovative and Engaging presentation.