Mark Wagner

Pretraining, data, and everything else necessary to build frontier AI.

I joined Inflection AI in 2022, before LLMs were widely available, because I had a conviction that LLMs could fundamentally transform how we think about technology products. As an early hire, I was responsible for data in all its forms. I scaled our data infrastructure to support pretraining-scale pipelines, built our data warehouse to support the product launch, did the early data science work for the company, and advanced the state-of-the-art in pretraining data curation.

SambaNova builds AI chips with a novel reconfigurable dataflow architecture. I worked mostly on the compiler and other parts of our software stack, with occasional forays into hardware and ML. I led the project to enable frontier-scale LLM training on the SambaNova stack, which led to our earliest commercial deployments.

For eight years, I worked to provide the data infrastructure LinkedIn needed to grow. LinkedIn's most impactful products ran on the analytics and machine learning infrastructure my team and I built.