AINLPLLM

AI-Based Text Anomaly Detection System

System that uses LangChain, LangGraph, and LLMs to analyze text and detect unusual patterns that may indicate fraudulent or suspicious activities, providing structured risk assessments.

GitHub

Architecture

Text input preprocessing & normalization

LangChain / LangGraph agent execution graph

LLM pattern analysis & anomaly detection engine

Structured explanation & risk scoring output

Results

Engineered an automated agent workflow using LangChain and LangGraph.

Generated contextual risk assessments and explanations for suspicious text patterns.

Challenges

Designing deterministic agent execution graphs with LangGraph

Context window limits and prompt structuring for subtle text anomalies

Minimizing false positive detections in domain-specific text

Future Improvements

Real-time streaming text analysisCustom vector embedding search integrationMulti-modal anomaly detection extension