ai ml2024

MedExtract

Medical data extraction platform with AI agent and MLOps

Tech Stack

PythonFastAPILangChainLangGraphAWS BedrockDynamoDBS3PostgreSQLpgvectorReactTypeScriptDockerAI AgentsRAGSystematic EvalsQuality GradingLangSmithBacktests

A production-grade medical data extraction and intelligent agent platform built for insurance underwriting. Combines automated extraction of unstructured medical documents into structured JSON with a conversational AI agent featuring 8 specialized tools (risk assessment, policy compliance, knowledge base search, etc.). Implements enterprise-level MLOps practices: systematic evaluation framework, complete observability with custom traces and LangSmith integration, LLM-based quality scoring, and regression testing with backtests. Supports dual interaction types (extraction and conversational) with field-level precision tracking and automated quality grading. Demonstrates production AI engineering maturity with comprehensive observability and quality assurance.

Screenshots

MedExtract screenshot 1
1 / 5

Key Highlights

Dual interaction types (extraction + conversational agent)

LangGraph ReAct agent with 8 specialized tools

Field-level precision tracking for quality assurance

LLM-based automated quality grading

LangSmith integration for complete observability

Regression testing with systematic backtests

Schema-driven architecture for consistency

Enterprise MLOps practices throughout

Metrics

8

Agent Tools

Extract + Chat

Dual Modes