Arghya Datta

About Arghya Datta

Arghya is an experienced researcher in the field of machine learning, artificial…

Experience

1. Ai Ml Scientist

   Amazon

   Sep 2021 — Present

   AI Scientist in the core research team at Alexa AI at Amazon working on Large Language Models (LLMs)\\\\nProjects:\\\\n• Improving Robustness of Models in presence of label ambiguity and noise: Led research and deployed mitigation techniques that reduces developer friction and online defects of LLMs in the presence of noisy labels.\\\\n• Reducing Model Instability in Large Language Models: Led R&D of novel metrics and LLM training methodology that improves model stability in predictions in discriminative LLMs.(ACL Findings 2023, https://arxiv.org/pdf/23•••••25.pdf)\\\\n• Optimizing offline test-sets that improve testing efficiency of LLMs before deployment by aligning test-sets with live traffic while simultaneously reducing size of test-sets by 50%.\\\\n\\\\nPreviously, I was a scientist in CBA (Customer Behavior Analytics) org focussing on Customer Segmentation and Customer Targeting for Amazon e-commerce website in US, IN, EU.\\\\n• Lead prototyping and deployment of deep neural network based propensity models that predicted customer purchase and browsing behavior on Amazon\'s e-commerce webpage. These models were then used in multi-arm bandits and online learning systems for efficient customer targeting initiatives.\\\\n• Built and deployed a novel scalable neural network layer that was able to rank high value actions (HVAs) for customers by leveraging propensity calibration in multi-label neural nets (published in Amazon Machine Learning Conference, 2022)\\\\n \\\\n\\\\n \\\\n \\\\n\\\\n \\\\n