CaptIA skopein

This repository presents a model-agnostic framework for the structured analysis of professional profile screenshots (e.g. LinkedIn profiles) using multimodal conversational language models. The framework defines: a fixed analysis pipeline (extraction, normalization, scoring, uncertainty handling), a standardized set of evaluation criteria, a normalized 0–100 scoring system with explicit weighting, a strict JSON output schema ensuring auditability and comparability. The system is designed as a decision-support tool, not an automated decision-making system. It explicitly separates visible information from uncertainty and does not infer sensitive attributes. A reference implementation is provided for reproduci bility purposes.