Sandesh Swamy

I am currently a Senior Applied Scientist at AWS AI Labs where I work on conversational agents, personalization, and Large Language Model (LLM) applications which make AWS customer experiences better. Prior to my stint at AWS, I was at Amazon Alexa from 2017 to 2021 where I worked on the in-house deep learning toolkit, traditional and neural-based Intent Classification and Slot Recognition models, and architected data pipelines and modeling pipelines for 100k skills. I also was part of the core set of initial contributors who started the Semantic Parsing based utterance recognition at Alexa. I have extensive experience working with (and deploying to scale) traditional Machine Learning models (max-ent models, linear chain CRFs), Deep learning models (LSTM-based models), and also Large Language Models (LLMs ~1B parameters). Before joining Amazon, I obtained my Master's graduate degree from The Ohio State University, Department of Computer Science in 2017. I was advised by Dr. Alan Ritter and Dr. Marie-Catherine de Marneffe (Still consider them the best advisers, ever! :)). My research interests include Natural Language Processing, Conversational agents, Semantic Parsing, Large Language Models (LLMs), Text Generation, Personalization, and Social Media data. I was previously the Teaching Associate for Introduction to Computer Programming in C++ for Engineers and Scientists.

I have significant programming experience in Python, Java and C(code samples and HackerRank profile can be found in the Navigation bar). I have dabbled with web programming sporadically. I also have extensive experience deploying real-world Machine Learrning systems which used frameworks such as MxNet, Pytorch, HuggingFace, and an in-house Amazon framework which have served millions of customers. During my time at Alexa, I have been a core contributor for the launch of traditional ML models for all Alexa skills, launching DNN-based tiny models for Alexa skills, release of Semantic Parsing based models for utterance recognition, international expansion of Alexa skills, and allowing customers to seamlessly request resource information using natural language on AWS' chat agent.

I am fascinated by the amount of information generated on Social Media. Analyzing Twitter data is a big interest of mine since it helps get a better understanding of the kind of things that people are really interested in and want to talk about.