hey, i am damir.
i build ai automation infrastructure at pragma: agent systems, data pipelines, and parsers that turn scattered documents, listings, and crm records into something you can actually query. it runs on open models, on hardware the client controls, so their data never leaves the building.
before that i built vetra ai alone, and victora ai, which avc group acquired and deployed internally.
i also run an independent research project on how the field validates single-cell potency scores. short version: the standard way cannot tell biological signal from sequencing depth. the long version is here.
problems i am interested in:
- whether the metrics a field trusts actually measure what they claim to measure
- agent infrastructure that runs on open models, on hardware the owner controls
- what happens to team structure when one person with models does the work of five
- getting all of this to people in central asia who have no access to it
before that i mentored 400+ engineers and founders at nfactorial, built upraze so kids would stop paying prep agencies thousands of dollars, captained the kazakhstan national taekwondo team, and interned at a bank while still in school.
almaty