Sandeep Krishnamurthy is an Engineering Director at Databricks, leading the MosaicAI Model Training Platform team that empowers enterprises to convert data into intelligent models. He is passionate about making foundation model training and inference faster, cheaper, and more accessible.

Sandeep holds a Master of Technology in Information Technology from International Institute of Information Technology-Bangalore. Before Databricks, he served as Senior Engineering Manager at Amazon SageMaker and led engineering teams at AWS AI, where he built the MXNet backend for Keras-MXNet.

Sandeep has deep expertise in deep learning systems, distributed training, and LLM inference optimization. He contributed to building MPT (Mosaic Pre-Trained) models and has been instrumental in democratizing foundation models for enterprises.

At Databricks, Sandeep led the team through the company's growth from $400M to $1.5B+ ARR over 3 years. He has worked on cutting-edge problems including efficient model serving, continuous batching, and optimizing inference for various accelerators.

Under his leadership, the MosaicAI platform has enabled enterprises to train and deploy state-of-the-art LLMs. His work on MPT-7B achieved training in 9.5 days with zero human intervention at approximately $200K cost, making commercial-grade LLMs accessible to more organizations.