Runjie Luo
Building Reliable AI Systems.
I design, evaluate, and investigate AI systems with a focus on agent reliability, observability, and human control.
Evidence·Evaluation·Observability·Human Control
Featured Skills
Evidence
Can AI explain what it did? Building systems where every claim traces to a source.
RAG
Grounding
Citation
Evaluation
Can we measure AI quality? Designing evaluation workflows for agent reliability.
Model Evaluation
AgentBench
Benchmarks
Observability
Can we trace AI behavior? Investigating why agent systems lose their own history.
Traces
Events
Audit Logs
Human Control
Where should humans decide? Designing human-in-the-loop gates that actually work.
HITL
Approval
Workflow
Featured Project
AuditFlow
AI-powered Audit Intelligence & Reliability Research
An AI system exploring how autonomous agents can produce verifiable decisions — through document intelligence, RAG, evidence grounding, and workflow observability.
Python
FastAPI
LangGraph
PGVector
Docker
DeepSeek
PostgreSQL
Redis
Latest Investigations
Forensic reads into why AI systems drift from their design.
Code Archaeology
Part 10"A single-implementation strategy ABC turned out to be a correct dependency-inversion boundary — the series' first positive finding."
Code Archaeology
Architecture
Abstraction
Code Archaeology
Part 9"The handover doc said the Review Agent wasn't wired to grounding — it was wired from the start, but to the wrong embedding model."
Code Archaeology
Grounding
Evidence
Code Archaeology
Part 8CRITICAL
"The knowledge base was indexed in 1024 dimensions; API queries ran in 1536 — two vector spaces in one system, and search could never work coherently."
Code Archaeology
Embedding
Vector Search