Tsinghua sociology master’s student with experience in AI products and experience research at Tencent, ByteDance, and Baidu—from model optimization to consumer products and 0-to-1 innovation. Rather than stopping at product specs, I turn ideas into working products myself.
With a background in sociology, I turn fuzzy ideas of good experience into measurable standards—and I like writing the code that turns ideas into working products.
I’m Shiqi Lei, a master’s student in Sociology at Tsinghua University. Over the past year, I have worked on AI products and experience research at Tencent Hunyuan, ByteDance’s TikTok and Flow, and Baidu ERNIE, focusing on translation, coding agents, and search.
I’m a classic ENFP: curious, energetic, and rarely content to sit still. I enjoy games, films, and books, but the question I keep returning to is how people relate to one another in society. Sociology gives me an extra layer of human-centered concern when I think about AI. I believe AGI will eventually make people more capable, while its risks deserve equal attention. As technology accelerates, we should also ask what kind of society it is creating. I’m excited to help shape new paradigms for human–AI interaction.
Tsinghua University · School of Social Sciences
M.A. in Sociology · Beijing · Aug 2024 – Jun 2027
Nanjing University · School of Social and Behavioral Sciences
B.A. in Sociology · Nanjing · Aug 2020 – Jun 2024
Ordered by relevance to AI. Each case covers the problem, what I did, and the outcome—open a card for the full story.
AI Product Engineer
Built and launched Tencent Hy Translate from 0 to 1 on top of the hy-mt2.0 model across multiple platforms, while establishing a translation evaluation framework and a bad-case-driven model improvement loop from scratch.
Model Data Platform Product Manager
Supported expert annotation for Seed Code post-training, improved the interactive trajectory annotation workflow, and connected trajectory generation, expert correction, and AI-assisted quality feedback into an end-to-end data production pipeline.
AI Search Product Manager
Used foundation-model capabilities to reshape the TikTok search experience, iterating systematically on intent detection, visual relevance, blended ranking, and citations, and helped launch AI search globally.
LLM Experience Research Assistant
Supported ERNIE Bot’s NLP track through user research and competitive insight, turning fuzzy user experience into model improvement strategies and data evaluation standards.
I do more than write product specs—I turn ideas into working products. Try the writing agent, browser translation extension, and mobile translation app below, or explore two reusable agent skills on GitHub.
A conversational agent that uncovers what you want to write, then learns your voice
A writing agent prototype that first clarifies your intent through conversation, then learns your writing habits to produce copy that sounds like you—not merely copy that reads correctly. The preset example below is fully interactive.
This interactive demo uses preset data and runs entirely in the browser—no API key required.
Instant selection translation, bilingual pages, floating controls, and a persistent sidebar
A Chrome / Edge translation extension with instant selection translation, paragraph-level bilingual webpages, floating controls, and a persistent sidebar. It also supports custom terminology and translation styles. Try the browser simulation below—no installation or API key required.
This interactive demo uses preset data and runs entirely in the browser—no API key required.
A mobile app for text, voice, and camera translation
A mobile translation app built on the hy-mt2.0 model, featuring streaming text translation, voice conversation, camera OCR translation, context memory, and multiple styles. The phone below uses sample data and is fully interactive.
This interactive demo uses sample data and runs entirely in the browser. Try the text, voice, and camera translation flows inside the phone.
Beyond building products, I package recurring workflows into reusable agent skills. Both projects are open source on GitHub.
Drop in daily screenshots, links, and files, then automatically organize them into a living personal knowledge base with iCloud capture and Feishu sync.
Read an existing codebase and reverse-engineer a review-ready product requirements document. It turns behavior scattered across routes, UI, defaults, and error handling into aligned requirements and acceptance criteria for PM, engineering, and QA.
Across research, data, design, and implementation, I can work hands-on throughout the AI product lifecycle.
If you’re looking for an AI product engineer who understands users and can build hands-on, I’d love to hear from you.