Internship Experience实习经历

AI Data Engineer — Beijing, ChinaAI 数据工程师 — 中国 · 北京

• Handled multimodal data collection, cleaning, and structured validation for basketball, pickleball, and horse trackwork (training) events, aligning video feeds, commentaries, and temporal events.负责篮球、匹克球及赛马晨操等项目的数据采集、清洗与结构化校验，深度整合赛事视频、解说文本、音频片段与时空事件标签。
• Prepared structured training sets for sports event recognition, highlight level grading, and training status classification of racehorses.提取赛事关键多模态特征，用于支撑运动事件识别、精彩程度评估以及马匹训练状态分类等算法的训练与精调。
• Maintained PostgreSQL schemas mapping sports events, video clips, generated commentary text, and model outputs to facilitate fast retrieval and evaluation.设计并维护 PostgreSQL 数据库，建立赛事、音视频片段、多语言解说词与模型预测结果之间的字段映射，确保高效数据检索与快速模型回溯。
• Constructed sports jargon dictionaries and automatic spelling-to-pronunciation error correction models, significantly boosting TTS pronunciation accuracy for player names and horse racing terms.构建体育垂直领域专用词典（覆盖球员姓名、赛马行话等专业术语），并参与设计自动纠错规则模型，使 TTS 模型朗读的专有名词准确率得到显著提升。
• Developed evaluation frameworks scoring generated commentary quality across fluency, emotive expressiveness, and style consistency, supplying algorithm teams with critical defect logs.建立多维度解说音频质量评估指标（如流畅度、情绪拟真度及风格符合度），对大模型生成内容进行质检打分，为 ASR/TTS 模型的迭代反馈关键缺陷数据。
• Designed prompt templates, golden evaluation sets, and error logs for automated highlight clipping, commentary generation, and audio synthesis pipelines.针对 AI 视频智能剪辑、赛事自动解说与配音生成，构建并迭代 Prompt 模版库、精选评估数据集以及典型错误案例集。