Markov packages 99 hours of engineering work into training data for desktop AI agents
The YC S26 startup paired 51 engineering tasks with gold files, rubrics, recordings and interaction traces across eight CAD software groups.
By Ryan Merket · Published
Primary source: X
Why it matters
AI agents need task environments and measurable outputs before they can move from browser chores into engineering software. Markov is building that training and evaluation layer around CAD.
Markov, the computer-use data startup founded by Dev Mandal (@DevvMandal) and Harish Ashok (@habril27), released a collection of human-performed CAD workflows on August 3rd for training and evaluating AI agents that operate engineering software through a graphical interface.
When you purchase through links in our articles, we may earn a small commission. This doesn’t affect our editorial independence.