TÜREL

Lab note

2026

Active

Local AI Workflows

Experiments with local models, practical automation, and private AI tooling.

Useful intelligence, kept close

Local AI Workflows examines where models running on personal and workstation-class hardware are already good enough to support real work. The emphasis is on bounded tasks: extraction, classification, drafting, transformation, and tool use with clear inputs and reviewable outputs.

Experiments compare model quality, memory requirements, latency, and energy use alongside the less visible cost of maintaining a dependable workflow.

Current questions

We are testing how small models can cooperate with deterministic tools, when retrieval improves results, and which interaction patterns make uncertainty clear without slowing routine work.

Topics

Local models · Automation · Private AI