Hi! I’m Udit Saxena.
I am a Machine Learning Engineer at Sumo Logic.
At Sumo, I build fast and approximate streaming clustering algorithms for text streams and time series models. I also work on running production ML workflows on Kubernetes and have experience with Airflow, AWS Sagemaker, Helm, and ArgoCD among other tools.
My research interests lie in the field of Machine Learning - specifically Natural Language Processing and sparse approaches to Deep Learning.
In my recent work A Unified Paths Perspective For Pruning at Initialization with Thomas Gebhart, we investigate how paths through neural networks determine performance and what this path structure tells us about how to optimally prune networks using the framework of Neural Tangent Kernels.
In the past, I’ve worked with Microsoft Research (at Montreal and Cambridge) on Active Learning, at Lexalytics on Graph Convolutional Networks for Text Classification, with Prof. Andrew McCallum on serving pre-trained Tensorflow models for the JVM as part of IESL, on a Pytorch implementation of End to End Memory Networks, as a Google Summer of Code intern, working with MLPACK (along with Ryan Curtin), as a Product Engineer at Sprinklr, and Adobe.
Check out my resume