Designing a voice-to-contract specialist that captures project details from real-time calls and generates structured, compliant contract drafts. Responsible for the conversation design, system prompts, and backend architecture connecting Twilio voice streams to AI extraction, validation, and document generation on Google Cloud. Currently in prototype phase (not yet demoable), focused on California home-improvement contracts; early tests suggest strong potential to eliminate manual intake steps and save contractors 10–15 hours per week on documentation.
Construction contractors were spending 15-20 hours per week manually transcribing voice notes into formal contracts and project documentation. This manual process was error-prone and prevented them from focusing on actual construction work.
• Construction terminology and requirements complexity\n• Need for high accuracy in legal documents\n• Human-in-the-loop validation required\n• Integration with existing contractor workflows\n• California compliance requirements
I designed and built an AI-powered workflow that processes voice recordings, extracts key project details, and generates structured contracts automatically. The system uses advanced speech-to-text and natural language processing to understand construction terminology and requirements. Built on Next.js with OpenAI's Whisper for transcription and GPT-4 for contract generation. Used Supabase for real-time data storage and user management. Implemented a multi-step validation process to ensure accuracy before final contract generation.
• 15+ hours per week saved on documentation\n• 95% reduction in transcription errors\n• 80% of contractors using within 3 months\n• Improved contract completion rates\n• Higher client satisfaction scores\n• Faster project turnaround times
As AI PM, I led the product strategy, user research, and technical implementation. Stack: Next.js (frontend) · OpenAI Whisper (speech-to-text) · GPT-4 (contract generation) · Supabase (database & auth) · Twilio (voice processing) · ElevenLabs (voice synthesis) · Cloud Run (API hosting) · Custom validation workflows
Would add more sophisticated audio preprocessing to handle noisy construction environments better. Also want to implement direct integration with popular contract management systems to eliminate manual export steps. Planning to expand to commercial projects and add multi-language support.