A small, independent AI/ML lab. We build open tooling and publish certified, quantized open-model checkpoints - held to a stated accuracy bar before they ship.

• Website: https://www.uist-labs.com
• Models on Hugging Face: https://huggingface.co/uist-labs

Certified quantization. We quantize open models and publish a checkpoint only after it clears a hard, stated accuracy gate against its full-precision baseline - measured on the identical evaluation harness, with the real deltas published alongside. A checkpoint that misses the bar doesn't ship. The name of the pipeline behind it says the idea plainly: an assay is the test that certifies a refined metal's purity.

Runs on constrained hardware. The checkpoints are weight-only NVFP4: they run via vLLM's Marlin FP4 kernel with no Blackwell GPU required, validated down to Ada-generation (sm_89) cards like the L4. Certified accuracy that travels to the hardware you already have, not the hardware you would have to buy.

• assay - the Apache-2.0 NVFP4 quantize -> benchmark -> gate -> publish pipeline behind every checkpoint we release.
• beautifulyze - render a piece of music into a picture a language model can read.
• cogito - an experimental framework for exploring autonomous AI cognition through recursive self-prompting.

• Security first. We don't trade security for convenience without saying so.
• Trust, but verify. Vendor claims are a starting point; the measured numbers are the product.
• Standard tools, honest docs. We prefer well-understood approaches, and we document the trade-offs, including what a release deliberately doesn't do.