Rongliang Sun
Research Engineer · Licon Studio · Shanghai, China · me@sommio.moe
About
Research Engineer at Licon Studio, working on multimodal data pipelines, video dataset processing, automated annotation, and AI-generated video quality evaluation. My background includes full-stack engineering, Linux infrastructure, and AI tooling.
Research Interests
- Video Understanding & Quality Assessment — video understanding and the evaluation of generated video quality.
- Multimodal Data Engineering — production, cleaning, annotation, and validation of image, video, and audio data.
- AI Evaluation & Tooling — model evaluation, automated assessment, and the engineering tooling around it.
Experience
Licon Studio
Research Engineer · Shanghai, China · July 2026 – Present
- Built structured data production pipelines for image and video generation workflows.
- Developed validation tools for machine-readable training data and dataset consistency.
- Processed and deduplicated large-scale video datasets (~135k retained clips, ~2 TiB), with per-record provenance and reproducible checks.
- Built multimodal annotation tools covering image, video, and audio analysis.
- Built evaluation workflows for AI-generated videos, including human-baseline comparisons and reproducible re-runs.
- Built auditing and usage statistics for model API calls.
Selected Work
Multimodal Annotation MCP
Python-based MCP tools for image, video, audio, and audiovisual analysis, letting AI agents annotate multimodal data through a single interface. Source on GitHub