Haizhou “Hydro” Li

I am a Senior AI Engineer working on LLM agents, multimodal AI, OCR/PDF intelligence, ASR/TTS systems, and production data pipelines. I am currently at Bosch Automotive Components in Shanghai, where I lead architecture and delivery for an intelligent cockpit in-vehicle Agent platform.

Before Bosch, I worked on AI Agent platforms, multimodal recruiting assessment, voice timbre evaluation, LLM-to-SQL applications, document intelligence, customer-service copilots, and large-scale multimodal corpus cleaning. My academic background is in Computer Science and Data Analytics: I earned an M.S. in Computer Science from Boston University and a B.S. in Economics & Data Analytics from Virginia Tech.

我是 高级 AI 工程师，主要方向包括大模型 Agent、多模态 AI、OCR/PDF 智能解析、ASR/TTS 系统与生产级数据 Pipeline。目前在 博世汽车部件上海团队 工作，负责智能座舱车机 Agent 平台的架构设计与落地交付。

在加入博世之前，我参与并主导过 AI Agent 平台、多模态招聘人像评估、声音音色评测、LLM-to-SQL 应用、文档智能解析、客服 Copilot，以及大规模多模态语料清洗系统。我的学术背景是计算机科学与数据分析：波士顿大学计算机科学硕士，弗吉尼亚理工大学经济与数据分析学士。

English CV 中文简历

Research and Engineering Interests

研究与工程方向

LLM Agent architecture, workflow orchestration, tool use, evaluation, and product delivery.
Multimodal AI systems including OCR, ASR, TTS, image captioning, portrait assessment, and data quality control.
NLP and retrieval systems, including LangChain, RAG, prompt engineering, topic modeling, and corpus construction.
Practical model adaptation and evaluation, including LoRA/QLoRA roadmaps, DPO planning, F1/AUC tracking, and regression workflows.

大模型 Agent 架构、工作流编排、工具调用、评测体系与产品化交付。
多模态 AI 系统，包括 OCR、ASR、TTS、图像 Caption、人像评估与数据质量控制。
NLP 与检索系统，包括 LangChain、RAG、Prompt Engineering、主题建模与语料构建。
模型适配与评测，包括 LoRA/QLoRA 路线、DPO 规划、F1/AUC 指标追踪与回归评测。

Experience

工作经历

Bosch Automotive Components, Senior AI Engineer, 03/2026 - Present. Leading architecture and platform delivery for intelligent cockpit in-vehicle Agent systems.
Shanghai Chicmax, Senior Algorithm Engineer, 04/2025 - 03/2026. Delivered an AI Agent platform for 8+ LLM use cases, plus multimodal recruiting assessment, voice evaluation, LLM-to-SQL product selection, and OCR document intelligence.
China Telecom Shanghai Ideal, Algorithm / Model Engineer, 06/2023 - 04/2025. Built customer-service copilots, PDF/OCR intelligence, and a multimodal corpus cleaning stack processing about 1 TB/day.

博世汽车部件，高级 AI 工程师，03/2026 - 至今。负责智能座舱车机 Agent 平台架构与交付。
上海上美股份有限公司，高级算法工程师，04/2025 - 03/2026。落地 AI Agent 平台，支撑 8+ 大模型场景，并建设多模态招聘评估、声音评测、LLM-to-SQL 选品与 OCR 文档智能系统。
中国电信上海理想信息产业有限公司，算法/模型工程师，06/2023 - 04/2025。建设客服 Copilot、PDF/OCR 智能解析，以及约 1TB/日规模的多模态语料清洗系统。

Selected Projects

精选项目

Global Knowledge Voice Translator: ASR + LLM + TTS localization product covering 100+ content sources and 5,000+ hours of overseas audio/video.
LLM Voice Assistant: mini-program voice assistant with persona selection, memory, vector retrieval, safety filters, and SoVITS-TTS multi-emotion speech.
AI Corpus Auto-Labeling & Cleaning: multimodal pipeline across text, image, audio, and video; improved usable data pass rate from 40% to 80%.
Deep Learning NLP Research: Boston University BIT Lab research using Flair, LDA, Fuzzy matching, and BERTopic. GitHub

全球知识语音转译小程序：ASR + LLM + TTS 本地化产品，覆盖 100+ 内容源与 5000+ 小时海外音视频。
大模型语音 AI 聊天助手：小程序端语音助手，包含人设、记忆、向量检索、安全过滤与 SoVITS-TTS 多情绪语音。
AI 语料自动打标及清洗：覆盖文本、图像、音频、视频的多模态 Pipeline，将可用语料通过率从 40% 提升到 80%。
深度学习 NLP 研究：波士顿大学 BIT Lab 研究项目，使用 Flair、LDA、Fuzzy matching 与 BERTopic。GitHub

Blog and Notes

博客与笔记

I keep a dedicated space for technical notes, project writeups, and longer-form thoughts. New posts can be added under _posts/ and will automatically appear on the Blog page.

我为技术笔记、项目复盘和长文保留了独立空间。以后只需要把文章放到 _posts/，内容就会自动出现在 Blog 页面。

Education

教育经历

Boston University, M.S. in Computer Science, 08/2021 - 05/2023.
Virginia Tech, B.S. in Economics & Data Analytics, 08/2017 - 12/2020.

波士顿大学，计算机科学硕士，08/2021 - 05/2023。
弗吉尼亚理工大学，经济与数据分析学士，08/2017 - 12/2020。

Haizhou "Hydro" Li