Munich🥨NLP: Featuring Legal NLP in Insurance

Transforming insurance with AI! Join Munich NLP in-person meetup for expert talks on LLM applications, followed by BBQ networking. 18:00 @ Warngauer Str. 30 (5min from U2 Untersbergstraße). All welcome!

Agenda:
18:00-18:30: Arrival & Networking,
18:30-18:40: Welcome & Intro,
18:40-19:20: First Technical Talk
Insurance Claim Processing with LLM-Powered Automation - Dr. Johann Hagerer
19:20-19:30: Short Break,
19:30-20:10: Second Technical Talk,
ContractExtract: Decoding Insurance Policies with LLMs - Baran Avinc
20:10 onwards: Barbecue Networking

Location:
Warngauer Str. 30, 81539 München @ Versicherungskammer Bayern
[Google Maps Link]

First Talk: Insurance Claim Processing with LLM-Powered Automation
Speaker - Dr. Johann Hagerer
This project delivers an intelligent insurance claim processing system that leverages Large Language Models to automate critical decision-making steps including completeness validation, policy extraction, and coverage determination. The system reduces claim processing time from days to hours while decreasing operational costs by 60-70%, eliminates human variability through standardized assessments that support regulatory compliance, and processes claims simultaneously without proportional staffing increases. Built on a modular three-stage LLM workflow engine with native Snowflake integration, the system features dynamic configuration management that enables Pydantic-based structured output validation for quality assurance and a hybrid intelligence design that combines LLM automation for routine decisions with human expert escalation for complex cases, representing a strategic transformation from reactive manual processing to proactive intelligent automation that scales with business growth.

ContractExtract: Decoding Insurance Policies with LLMs
Speaker - Baran Avinc
At Sparkasse Versicherungsmanager we developed ContractExtract, as part of which we offer an AI solution that automatically analyzes insurance contracts through a simple document upload. For the development, we chose a structured approach to key information extraction from insurance contract PDFs and went through an intensive evaluation phase in which we systematically compared different architectural approaches and LLMs. The central challenge lay in optimizing the system with regard to three competing factors: extraction accuracy, response times, and the technical complexity of the overall architecture. In numerous tests, we searched for the optimal balance of these parameters.

Thanks to our sponsor Sparkasse, VersicherungsKammer and Inverso for hosting the event.

Organizer:
Munich NLP,
Dr. Johann Hagerer (Inverso GmbH),
Victor Tremols, Head of Data Analytics (AI/BI). (Versicherungskammer Bayern),