Contact person: Business Client

Phone:Show

Email:Show

Location: Phoenix, United States

Budget: Recommended by industry experts

Time to start: As soon as possible

Project description:
"I want to turn my deep knowledge of residential mortgage lending into an always-available assistant powered by a custom GPT. The core model must confidently answer questions about products, guidelines, current rates, industry trends and anything a borrower, realtor or loan officer might ask.

Phase 1 – Build the brain
• Fine-tune or expertly prompt-engineer a GPT that draws from compliant, up-to-date mortgage data.
• Expose it through the OpenAI API so I can hit one endpoint from anywhere.

Phase 2 – Wire in automations
The same GPT needs to trigger workflows in both [login to view URL] and Zapier. Typical flows will populate lead forms, run payment-estimate calculators and push results back to users in real time. Webhooks, custom functions and OAuth setup are expected here.

Phase 3 – Deploy everywhere
The first roll-out is intentionally broad: website widget, social-media chatbots and a cross-platform mobile app (React Native or Flutter are both fine). A single codebase with responsive UI will keep maintenance low, and the GPT endpoint should be reusable across all surfaces.

Deliverables
1. GPT configuration (prompts, system message, or fine-tuned model) with clear update instructions
2. Secure API layer with authentication and rate limiting
3. Two sample automation scenarios in [login to view URL] and Zapier, fully tested
4. Starter front-end implementations:
– Embeddable JavaScript chat widget for any website
– Social-media bot script ready for Facebook/Instagram Direct
– Cross-platform mobile shell that consumes the same endpoint
5. Short hand-off guide covering environment variables, deployment steps and future scaling tips

Acceptance Criteria
• Mortgage answers are accurate, cite sources when asked and never expose private keys.
• Automations fire within three seconds and return calculated outputs correctly.
• All code runs locally with a single README command and deploys to my chosen cloud account without modification.

If this sounds like your wheelhouse, let’s build the smartest mortgage assistant on the market." (client-provided description)