Prototype
Manufacturing Knowledge RAG
Question
How can manufacturing documents be retrieved with reliable source traceability?
Approach
Chunking, embedding, vector retrieval, citation-aware generation.
Tools
FastAPIQdrantBGELangChain
AI Engineering
AI Lab
Experiments in manufacturing knowledge RAG, agent workflows, semiconductor analytics, and AI engineering.
RAG Systems
Agent Workflows
Semiconductor Analytics
Evaluation
Experiments
Prototype systems for manufacturing AI workflows.
Research
LangGraph Agent Workflow
Question
How can multi-step manufacturing tasks be decomposed and routed by agents?
Approach
Planner, tool calling, workflow routing, human review.
Tools
LangGraphLangChainPython
Case Study
SECOM Yield Prediction
Question
How can high-dimensional semiconductor process data support defect risk modeling?
Approach
Feature selection, imbalance handling, LightGBM, SHAP analysis.
Tools
Pandasscikit-learnLightGBMOptuna
In Progress
Semiconductor Defect Analytics
Question
How can visual and process signals support semiconductor quality analysis?
Approach
Defect pattern analysis, process feature correlation, model evaluation.
Tools
PythonPyTorchOpenCV
Methods
Evaluation before automation.
The lab focuses on reliability constraints that matter in production settings: retrieval quality, source traceability, agent evaluation, and human review.
Retrieval Quality
<built-in method copy of dict object at 0x7224e658a600>
Agent Evaluation
<built-in method copy of dict object at 0x7224e5ce5b00>
Source Traceability
<built-in method copy of dict object at 0x7224e65fa700>
Human-in-the-loop Review
<built-in method copy of dict object at 0x7224e5335dc0>