Aug 2023 – Oct 2023 · 2 mos

Feb 2023 – Aug 2023 · 6 mos

• Performed extensive feature engineering to capture wide range of signals and built a unified sales recut model to predict sales for each size of each product outperforming the baseline by 10%.
• Improved the robustness of the scraper while running multiple instances in parallel by 30% by refactoring the scraper architecture and rate-limiting the scraper further processed data using PySpark
• Built an internal platform to annotate the data and improve recommendations

Aug 2022 – Dec 2022 · 4 mos · Seattle, Washington, United States · On-site

• Team: Prime Video
• Project: Reinforcement learning based Recommender system
• Gained improvement of 7% HitRate@10, with Reinforcement Learning based model for title level recommendation. This work was published and presented at Amazon Machine Learning Conference (AMLC ’22 RL workshop)
• Increased the performance of the North-South Carousel Ranker by 4.7%, in terms of HitRate@3, using a Transformer based post-processing technique.

May 2022 – Aug 2022 · 3 mos · Seattle, Washington, United States

• Team: Siri Perception
• Project: Noise robust acoustic feature extractor
• Trained Feature Extractor on 84K utterences of Speech-command dataset which could be jointly optimized with a number of downstream tasks
• Performed noise aware training of feature extractor that denoise various acoustic condition improving downstream accuracy by 5%
• Shrank the model parameters by 50% with negligible degradation in performance
• Researched and Explored self-distillation based teacher-student model and performed base experiments with it

Dec 2020 – Apr 2021 · 4 mos · India

• TITLE - Comparison of INSAT-3D/3DR precipitation estimates with reference to global satellite rainfall data over specific climatic regimes.
• Tasks performed:
• Constructed calibration and validation algorithm for comparing image datasets from multiple satellites (200+ GB)
• Created an amalgamation of 3 interpolations (nearest-neighbor, bicubic, and bilinear) to avoid pixelation
• Programming Language: Python
• Tools/Libraries: Pandas, Matplotlib, Scipy, h5py, geopandas, etc.

Aug 2020 – May 2021 · 9 mos

• Role:
• 1) Mentored a team of 5 juniors to develop and solve various problems using Machine Learning and Deep Learning.
• 2) Coordinated a seminar on Introduction to AI along with the junior team. Also, delivered various webinars and workshops on Data Version Control, Intro. to Machine Learning, and etc.

Oct 2019 – Aug 2020 · 10 mos

• Role:-
• 1) Delivering Lectures and seminars on Machine learning concepts.
• 2) Managing the Events organized.
• 3) Learned new technical concepts from my senior team members.

Apr 2020 – Jun 2020 · 2 mos

• Accelerated data ingestion speed by 80% by indexing SQL tables and refactoring SQL queries
• Increased code coverage of the system by 35% by writing unit and integration test cases using Pytest
• Virtualized the servers using Docker for the test environments and dev-environments needs
• Programming Language, Tools, Libraries, and skills used
• Python, Pandas, Numpy, JIRA, GIT, GitHub, Docker, CI/CD, Pytest, and Object Oriented Concepts