Engineer by craft, storyteller on stage.
I'm Saumya — I build the unglamorous machinery that makes machine learning trustworthy, and I love standing in front of a room explaining why it matters.
I'm a Senior Machine Learning Engineer at Datamics, where I help teams take MLOps and LLM systems from prototype into production — experiment tracking, model lifecycle, and the observability that keeps models honest once they're live.
Before Datamics I was at Verizon. I'm an AWS Solutions Architect and an Informatics graduate of the Technical University of Munich. The two sides feed each other — research sharpens the engineering, and production keeps the research honest.
Somewhere along the way I discovered I love the stage. In my talks and my writing, I try to take the hardest parts of modern ML — LLM observability, evaluation, the reliability nobody tweets about — and make them land. That's the throughline: build it well, then help others build it too.
Clarity over cleverness
The best system — and the best talk — is the one people actually understand. I optimize for the moment it clicks.
Observable by default
If you can't see it, you can't trust it. I build ML the way I'd want to debug it at 2am.
Teach what I learn
Writing and speaking aren't a side quest — explaining an idea is how I know I truly understand it.