I joined TikTok in March '24 in the Monetization GenAI (Ads) team. Most of my work has revolved around curating, cleaning and pre/post processing internal business data for both continual pretraining and supervised finetuning of LLMs. I've leveraged these trained models to solve several downstream use cases, spanning a number of TikTok Monetization businesses.I work as a Senior Data & Applied Scientist in the Azure Machine Learning team at Microsoft, and my focus has been on Natural Language Processing, and Data Science in general. Most recently I've been involved in projects focused on providing enterprise solutions to our customers leveraging the power of the very recent LLMs, including OpenAI as well as OSS models like LLaMA. Previously I've been involved in extensive experimentation and benchmarking of the SOTA LLMs and the delivery of AutoML for NLP product offering (now in General Availability). I've also led many interactions with customers and helped many new and existing customers leverage as well as get the most benefit from our Automated Machine Learning offering.Areas of expertise include Deep Neural Networks, Large Language Models, PyTorch based model implementation, as well as classical machine learning.I have also successfully completed graduate courses like Deep Learning, Statistical Machine Learning, Applied Statistics and Experimental Design, Introduction to Statistics and Probability, Information Visualization, and Database Management at the University of Washington, Seattle with a cumulative GPA of 3.92 I've been a Research Assistant at the Institute for Health Metrics and Evaluation, where I worked on estimating the burden of disease averted by personal health care in the US, where burden is measured by disability-adjusted-life-years and personal health care by the healthcare quality and access index.I previously worked in Algorithmic Trading at Goldman Sachs, for 5 years where I analyzed market data, in the form of time series data, to calibrate signals for trading algorithms and worked on numerous regression problems to determine the optimal tolerance bands for trading. I've played a multitude of roles, from quantitative research on volume profiles to being a lead developer for highly time sensitive trading algorithms in an event-driven programming language. I've led high profile projects with 10+ people, collaborating with tech, sales, researchers and clients. I was also the Product Owner for the benchmark trading strategies and the expert on many controls that helped regulate the trading algorithms.