LLM Chatbot on $0 Budget | Home Insurance Democratization

The Human Story: “After working 26-hour days to buy your first home, the last thing you need is decision fatigue about insurance policies written in legal hieroglyphics.”

The Challenge: Housality (pre-seed startup) wanted to build an insurance policy analyzer chatbot. Problem: No public datasets of real insurance policies exist, and every API costs money they didn’t have.

The Hustler’s Solution: I became a master of deception—called insurance companies pretending to be a customer, collected sample policies, then built a synthetic data pipeline to create 1,000 realistic policies per minute. All using free resources.

The Technical Marvel:

• LLM Fine-tuning: Tested 7 different models across 20+ metrics
• Synthetic Data Generation: Created realistic policy variations from templates
• RAG Infrastructure: Built knowledge base without spending a dime
• Economic Engineering: Negotiated 50% discount on MLOps platform ($50/month vs $100)

The Impact:

• Business: First profitable component of their property evaluation service
• Human: Takes decision fatigue away from exhausted new homeowners
• Technical: 13% boost in LLM faithfulness with zero paid dependencies

Tech Stack: Multiple LLMs, RAG, Synthetic Data Generation, MLOps “When you have no budget, creativity becomes your superpower.”