Viviane Alencar

I’m a technical documentation and linguistic quality assurance professional focused on AI content quality and LLM evaluation. With 12+ years in complex engineering software, I identify ambiguity, inconsistency, and structural failures where precision matters most.

In parallel, I built hands-on research depth through a B.Sc. in Applied Mathematics and Physics, with a thesis on CLIP-like embedding space analysis and zero-shot learning. I trained and benchmarked 17 custom neural network models against an OpenAI ViT baseline, studying how batch size, dimensionality, and dataset scale affect image-text embedding geometry.

Professionally, I advise on visual identity and design decisions for technical audiences, strengthen scannability and reading flow, and provide editorial reviews across articles and technical publications. I bring the same rigor to AI-related content—evaluating clarity, coherence, and failure modes so outputs meet professional communication standards.

Previously, I delivered localization and documentation quality end-to-end at Autodesk and Delcam: maintaining UI/help content, building internal terminology infrastructures on Confluence, and coordinating with development teams through Jira. I also contributed to metrology documentation by designing controlled experiments, documenting results with traceability, and turning findings into cleaner technical specifications.