Technologies: Python | Open-Weight LLMs | Data Pipeline Architecture | HCI

---

Processing multimodal behavioral telemetry (such as audio and spatial motion) from XR sessions is highly resource-intensive.

To systematically analyze complete interview corpuses without sacrificing analytical rigor, I built a novel Human-in-the-Loop (HITL) methodology utilizing a confidence-weighted ensemble of open-weight Large Language Models (LLMs). This pipeline automatically extracts granular usability insights and thematic codes, drastically reducing researcher thematic analysis time while maintaining high inter-rater reliability.