Jincheng Lyu

Ann Arbor, MI · University of Michigan

Jincheng Lyu

Building grounded AI systems.

ML & Data Science graduate student. I build knowledge-grounded systems for large language models: retrieval over billion-scale knowledge graphs, evidence and provenance layers, and the full-stack engineering around them.

Resume (PDF) Email GitHub LinkedIn

M.Eng. Data Science & ML, UMichB.S. CS, Penn State · Schreyer HonorsAIED 2025

About

I am an M.Eng. student in Data Science & Machine Learning at the University of Michigan (expected Apr 2027), after completing my B.S. in Computer Science at Penn State with a Schreyer Honors thesis on intent recognition in guiding dialogue.

My research work at the Jieliu Lab focuses on biomedical knowledge graphs: serving 38M+ scientific articles and 3.2M+ entities through retrieval, evidence, and provenance APIs that keep LLM agents grounded and auditable.

I care about LLM reliability, knowledge-grounded generation, and systems that make models verifiable. I also ship production web apps end-to-end, from PostgreSQL schemas to React frontends.

Experience

Research Assistant · Jieliu Lab, Dept. of Computational Medicine & Bioinformatics, UMich May 2026 – Present
  • Built the interactive frontend for a biological knowledge graph platform (38M+ articles, 3.2M+ entities) with React, TypeScript, Vite, Three.js, and react-force-graph-3d, including a 3D force-directed cluster overview and a 2D subgraph explorer with 3-hop BFS-distance layout.
  • Maintained the production FastAPI backend and PostgreSQL database: API endpoints, persistent conversation state, streaming improvements, and fixes for migration and serialization bugs causing 500s.
  • Built and deployed a GPU resource tracking dashboard with calendar heatmap and drag-and-drop allocation UI for the compute cluster.
Research Assistant · National AI Institute for Exceptional Education (NIEAE) Sep 2024 – Present
  • Built ASR and speaker classification pipelines for automated assessment of parent–child interactions in autism interventions.
  • Contributed data quality control and AI output validation to the AIED 2025 paper as third author.
Full Stack Developer · RadicalAI (startup) May – Aug 2024
  • Developed core product features for an early-stage startup using React, contributing directly to the production codebase.
Honors Thesis Researcher · Schreyer Honors College, Penn State Aug 2024 – Dec 2025
  • Developed a dialogue classification system for NLP research with a faculty advisor; produced the honors thesis: GUIDE: A Dataset and Study of Context for Intent Recognition in Asymmetric Guiding Dialogue.
Learning Assistant, Data Structures · Penn State University Jan – Dec 2023
  • Supported undergraduate students in Data Structures coursework through guided problem-solving and concept reinforcement.

Projects

GLKB

Research

Biomedical knowledge graph platform. Semantic search, entity resolution, and evidence-backed answers over 38M+ articles with full provenance tracking.

Python · FastAPI · Neo4j · PostgreSQL · pgvector · LLM agents · MCP

My Car Story

Product

A living log of my Honda Civic rally-inspired build: every part, every job, every drive. Full-stack app with auth, rich content, and media. my-car-story.com

React · Node.js · Express · PostgreSQL · Supabase

Lab Scheduler

Tooling

Team scheduling platform for a research lab. Migrated a Git-backed JSON task system to PostgreSQL, adding role-based access and real-time state.

FastAPI · PostgreSQL · React · Docker

MAI-T1D

Research

Governance knowledge graph dashboard for type-1 diabetes AI research: provenance, coverage matrices, and blast-radius analysis over structured knowledge.

React · FastAPI · Neo4j · AWS

Hermes Traffic Light

Hardware

A physical AI agent status indicator built on an ESP32-C3 microcontroller, controlled through a Tailscale-bridged HTTP API. github.com/BruceBest/agent-light

ESP32-C3 · C++ · HTTP API · Tailscale

LUMI LINK · Online Forum

Full-stack

WebSocket chat application with database-interface logic deployed on AWS EC2, plus a server-side Java application using HTTP for user-database communication.

Java · WebSocket · HTTP · AWS EC2

Publications

ASD-HI: A Parent-Child Interaction Dataset for Automated Assessment of Home Intervention

Zhaohui Li, Yusuf Akemoglu, Jincheng Lyu (third author), Qingxiao Zheng, et al.

In Artificial Intelligence in Education (AIED 2025), Lecture Notes in Artificial Intelligence, vol. 15877, pp. 48–62. Springer, 2025.

Education

M.Eng. Data Science & Machine Learning · University of Michigan, Ann Arbor Jan 2026 – Apr 2027
  • Focus on machine learning, NLP, and large-scale systems.
B.S. Computer Science, Minor in Mathematics · Penn State University (Schreyer Honors College) 2021 – 2025
  • Honors thesis on intent recognition in guiding dialogue, advised by Prof. Rebecca J. Passonneau.
  • Schreyer Honors College · NSLS.

Skills

Languages: Python, TypeScript, JavaScript, Java, C, C++, SQL, Verilog.

ML / AI: LLM integration, Fine-tuning, RAG, Retrieval, Knowledge graphs, MCP servers, Hermes Agent, Claude Code, OpenAI API.

Backend: FastAPI, Node.js, Express, PostgreSQL, Neo4j, pgvector, Redis.

Frontend: React, Vite, Three.js, HTML/CSS.

Infra: Docker, GitHub Actions, AWS EC2, Supabase, Cloudflare Tunnel, Linux.

Spoken: Mandarin (Native), English (Fluent).

Contact

I’m looking for full-time roles in ML, data engineering, and AI systems starting Apr 2027. Research collaborations welcome too.

Download Resume brulyu@umich.edu github.com/BruceBest linkedin.com/in/jinchenglyu