Xiao L.

Key Skills

Software

No software listed

Top Subject Matter

Multimodal educational AI training data (Q&A and teaching scenarios)

Educational digital-human and video content for AI fine-tuning

Educational images (teaching/poster materials) for multimodal training

Top Data Types

Text

Video

Image

Document

Top Task Types

Question Answering

Fine-tuning

Bounding Box

Object Detection

Text Generation

Classification

Freelancer Overview

Product Manager—defined and executed multimodal educational data labeling pipelines for Q&A texts and related teaching m. Brings 4+ years of professional experience across complex professional workflows, research, and quality-focused execution. Core strengths include Seedance2.0, GPT-image-2, and Large Language Models. Education includes Bachelor of Science, Sichuan Technology and Business College (2022) and Bachelor of Science, Mianyang City University (2024). AI-training focus includes data types such as Text, Video, and Image and labeling workflows including Question Answering, Fine-tuning, and Bounding Box.

Labeling Experience

Product Manager - Sichuan Shengfu Technology Co., Ltd.

ImageBounding BoxClassification

As a Product Manager, you independently created labeling specifications, taxonomies, and quality inspection standards for multiple industrial and archive data types used in an archive management system and related AI modules. You led full-lifecycle processing of raw materials, including collection, cleaning, desensitization, deduplication, and invalid-data filtering, and organized unified data formats for production. You used GPT and AI image/video generation tools to expand the raw-material library and coordinated with R&D teams to report defects, refine rules, and produce training datasets for generative model iteration. • Authored data labeling operation manuals, taxonomies, and inspection standards across structured texts, product images, and equipment parameters • Built and maintained an online material data ledger with cross-sampling inspections and rework • Directed preprocessing pipelines and refined task design to improve consistency and accuracy • Partnered with R&D to optimize labeling rules and generate training datasets for AI model iteration

2025 - 2026

AI data preparation and labeling pipeline contributor—expanded industrial multimodal training material library and supported iterative improvements.

VideoText Generation

Mass-produced industrial product raw materials and promotional assets to expand the pool of data pending labeling. Used AI generation tools to create industrial product copywriting and short videos as part of the training-data library. Coordinated with R&D to report labeled-data defects and iteratively refine labeling rules based on observed issues. • AI-assisted industrial product copywriting • Promotional product images and short-video material generation • Iterative optimization of labeling rules with R&D • Defect reporting and continuous dataset quality improvement

2025 - 2026

Product/AI data operations—prepared industrial image assets with AI-assisted pre-labeling and quality control for model training.

ImageObject Detection

Performed image and equipment-related annotation preparation to support industrial generative model training. Applied pre-labeling using AI image/video generation tools and completed initial steps such as object detection for pending labels. Maintained consistency through ledger-based tracking and periodic cross-sampling inspections. • Pre-labeling for image object detection • Preparation of real-shot product photos for labeling • Online material ledger maintenance and auditing • Cross-sampling inspection and rework for label quality

2025 - 2026

Labeling/AI data spec owner—authored labeling rules and processed structured text data for industrial AI training datasets.

Document

Developed labeling specifications, label taxonomies, and quality inspection standards for structured archive text data. Led full lifecycle processing including collection, cleaning, desensitization, deduplication, and invalid-data filtering. Generated standardized labeling operation manuals to ensure consistent execution across data pipelines. • Structured archive text labeling • Text entity extraction as a pre-labeling step • Unified data formats and refined task decomposition • Quality inspections via cross-sampling and rework of erroneous labels

2025 - 2026

Product Manager - Hefei Wanbo Zhengxin Education Technology Group Co., Ltd.

Image

As a Product Manager, you owned product planning and execution for education technology solutions including AI Q&A Rooms, digital human teaching platforms, and PD teaching systems. You translated teaching scenario needs into clear labeling and data-delivery requirements, then defined workflows and quality standards to support downstream development. You collaborated with cross-functional stakeholders to ensure consistent, high-quality production and release of labeled content and datasets for model fine-tuning and product iteration. • Defined labeling workflows and unified quality inspection/error-tolerance standards • Coordinated data demand analysis for teachers and students across multiple teaching scenarios • Managed preprocessing, organization, batch review, and deviation correction for multimodal materials • Used AIGC tools and LLMs to generate raw materials and supported dataset quality assurance

2023 - 2025

Show more

Education

Work History