AI Engineer previously at CERN, EQT, Deutsche Bank

Eirik Steen  • Peerlist

We built Kvasir, a system for parallel agents to analyze data, run models, and quickly iterate on experiments based on context graphs that track data lineage.&nbsp;We built it as ML engineers who felt existing tools weren’t good enough for real-world projects we have done. Most analysis agents are notebook-centric and don’t scale beyond simple projects, and coding agents don’t understand the data. Managing experiments, runs, and iterating on results tend to be neglected.Upload your files and give a project description like “I want to detect anomalies in this heartrate time series” or “I want to benchmark speech-to-text models from Hugging Face on this data” and parallel agents will analyze the data, generate e-charts, build processing/modeling pipelines, run experiments, and iterate on the results for as long as needed.