Yale University
Biostatistics Data Science
Advanced coursework in statistical methods, machine learning, data analysis, and computational biology. Research focus on NLP and knowledge graphs in genomics.
Gerstein Lab ResearchNLP & Genomics
About
Builder in transition
I'm a self-taught AI/LLM engineer in progress, currently transitioning into the field by studying GenAI and building real-world LLM apps.
My academic background spans Information Mgt, Finance, and Biostat, but growing fascination with LLM led me to pivot toward AI engineering. I began this journey in 2025, and since then I've been focused on studying how modern large language model systems work.
Through self-study and hands-on projects, I've been exploring the end-to-end LLM stack — from model fundamentals to inference optimization and downstream applications such as RAG pipelines, agentic workflows, and AI-powered search systems.
I believe in the AI era, the most valuable asset is not a single background, but curiosity, discipline, and the ability to fast-learn and adapt.
Education
Academic path
Capital University of Economics and Business
Information Management and Information Systems (Finance.)
Comprehensive coursework in Java, Python, Database Systems, Web Design, and System Design. Strong foundation in information systems, data management, and software development.
Academic Excellence ScholarshipMerit StudentInformation Systems
Experience
Applied work
ETH | Machine Learning Engineer Intern (Nework, CA)
Agentic Multimodal Search System
Built an Agentic Multimodal Search System POC to improve search coverage and relevance, supporting hybrid retrieval and temporal queries. Developed a multimedia indexing pipeline and an LLM query understanding module for NL-to-DSL parsing, fine-tuning domain-specific LLMs to optimize retrieval and server-side performance.
Yale Gerstein Lab | Research Assistant
BioGraphRAG Biomedical Retrieval System
Built a GraphRAG-style biomedical mechanism retrieval system with DAG reasoning constraints to reduce semantic drift. Designed a comprehensive retrieval pipeline integrating Neo4j multi-hop search, semantic index construction, and FAISS reranking to generate traceable mechanism explanations.
Skills
Working toolkit
Languages
PythonRPyTorchTypeScriptSQLJava
Frameworks & tools
LangChainLLaMA FactoryElasticsearchDockerFastAPILinuxGitAWS
AI systems
RAGAgentic workflowsHybrid retrievalFine-tuningLLM evaluationGraphRAG
Product instincts
ObservabilityCost routingLatency tuningStructured UXTechnical writing
Projects
Agent project showcase
3 of 5 projects
AI-Native Project
Agents and workspaces built around retrieval, reasoning, and operational AI workflows.
Multimodal search agent
A text-to-any retrieval agent that can search across different data formats from natural language.
Multimodal retrievalAgentSearch
View repository
EvoSearchAgent
An agent that improves how it searches through your knowledge database over repeated use.
Knowledge baseAdaptive searchAgent
View repository
Medicare_agent
A structured clinical interview system for collecting patient context and supporting safer medical reasoning.
Clinical AIInterview systemFastAPI
View repository3 of 4 projects
Make me more productive (AI-Powered)
Personal utilities that turn messy inputs into searchable, usable learning and work systems.
SnapNote
A productivity utility that turns quick screenshots into structured AI knowledge notes.
ScreenshotsNotesKnowledge
View repository
DeepFocus
A focused learning video website designed to reduce distraction while watching and studying.
LearningFocusVideo
View repository
Academic Research Agent
A full-stack academic research agent for multi-session research assistance and knowledge workflows.
ResearchMulti-sessionFull stack
View repository3 projects
Just for fun (AI-Powered)
Experiments where AI is used as a playful interface, creator tool, or narrative system.
Life.OS
An interactive narrative experience wrapped in a cinematic interface.
NarrativeInteractiveNext.js
View repository
AI Co-Creator
A RedNote AI co-creation tool for drafting and shaping creator content.
RedNoteCreator toolContent AI
View repository
GSEO
An AI search content optimization tool for RedNote content creators.
Search contentOptimizationCreator AI
View repositoryBlog
All postsLatest notes
AI Engineering / Programming Languages
对 AI 时代编程语言的感悟
在现在的 AI 时代，如果还抱着过去那种“死磕一门语言、做某个语言的垂直专家”的旧观念不放，就是所谓在用战术上的勤奋来掩盖战略上的懒惰。
Knowledge Management / Learning Systems
关于知识管理的一点反思
最近一段时间一直在快速学习大量新的技术知识。AI、LLM、Agent、系统设计、云计算、软件工程，每天接触很多新的概念。我习惯把学习内容整理到 Notion 里面，通过建立不同层级的页面，把知识按照主题进行分类。
AI Safety / Agent
Recursive Self-Improvement in AI：递归自我改进到底是什么？
讨论下 Anthropic 关于 RSI，也就是 Recursive SelfImprovement 的研究。
Podcast
All podcastsAudio notes
Coming soon
How I learn AI engineering in public
A placeholder episode about turning notes, projects, and deployment mistakes into a portfolio that can be reviewed by real engineering teams.
AI engineeringPortfolio
42 minComing soon
From blog notes to searchable conversations
A future discussion on making long-form learning material easier to search, summarize, and revisit through retrieval-augmented interfaces.
RAGSearch
36 minComing soon
Building personal software with production habits
A draft episode about shipping small public tools with clear version control, scoped releases, and realistic infrastructure choices.
SystemsVercel
51 minContact
Send a message
Let's talk about AI systems, research, or portfolio work.
The form opens your email client with a prepared message. Direct email works too.
chengshuo.dai23@gmail.com