A Rīga data R&D house. We build AI systems that survive production — and we publish the receipts.
primitives — six runtime QA mechanisms from the production enrichment engine behind unbin.io: per-field truth discovery with learned source reputation, confidence calibration with an ECE drift alarm, deterministic citation self-verification, completeness-regression and dead-seam alarms, a deterministic prompt-injection envelope, and an uncertainty-driven fetch policy our own 91-SKU benchmark refused — published together with the fixture that refused it.
Every number in that repository comes from production telemetry or from the benchmark that judged the mechanism. Published source, all rights reserved — for reading and evaluation.
Long-form: bargo.lv/notes — each primitive's mechanism, the production incidents behind it, and the deletion log.
Working with us: bargo.lv · four engagements a year · arnis@bargo.lv