Cyrus OS — Personal AI Desktop Assistant (Independent)
Built an AI desktop assistant and prompt/LLM evaluation workflow to improve response quality and routing across multiple models. Worked on structured assessment and rubric-style evaluation for generated AI outputs in the context of the assistant’s features. Designed components that support training-like iteration by measuring and refining model behavior over time. • Prompt design and multi-model LLM routing • AI response evaluation using structured rubric assessment • Security and audit logging to track assistant interactions • Vector-memory integration to support context-aware responses