Inspiration
Inspiration
In my mortgage advisory work, I could not always respond to prospects quickly. Reviewing and answering messages took time, including conversations with people whose needs did not match the services I offered or who stopped responding.
I wanted a practical way to organize those conversations, gather the information needed for an initial review, and keep the context available when I stepped in personally. NYVARA Hipotecas grew out of that everyday business need.
What I learned
I learned how to adapt technology to my business needs. Working with GitHub, Vercel, Twilio, and AI tools helped me understand how a conversation, stored information, and a web dashboard can work together.
I also learned that a successful deployment is only the beginning. Testing real interactions revealed problems that needed attention, from interpreting amounts and simple “yes” answers to making sure a notification actually reached the advisor.
How I built it
My role was to bridge technology and everyday business problems. I defined the questions, tested conversations, identified mistakes, and decided how the workflow should behave. I wanted a solution tailored to my needs, without unnecessary buttons.
With assistance from ChatGPT, Work, and Codex, I developed and tested a prototype that connects WhatsApp conversations to a protected Command Center. It gathers information for home purchases and mortgage refinancing, preserves conversation context, and lets the advisor review each prospect’s profile.
When a prospect requests human assistance, NYVARA pauses its automated conversation and attempts to notify the advisor through WhatsApp. The advisor can take over from the panel and later return the conversation to NYVARA.
Challenges
One of my biggest personal challenges was becoming comfortable with technical terminology and unfamiliar tools.
The technical challenges included understanding answers written in everyday language, distinguishing a property’s price from the down payment, handling answers that could not be interpreted, and managing the transition between AI and human attention.
Repeated testing helped us improve the experience. For example, after two unsuccessful attempts to answer a question, NYVARA offers human assistance and asks for consent before registering that request.
What I am proud of
My biggest achievement is completing this stage of the project through an iterative process with AI tools. I brought the business experience and made the decisions; the tools helped me turn those requirements into a working prototype.
We tested the journey from a WhatsApp conversation to a completed profile, an advisor notification, human takeover, and a return to NYVARA with the previous context preserved.
What comes next
The current version is a proof of concept tested with fictional data through the Twilio WhatsApp Sandbox. A completed profile is not a credit approval, and an assistance request does not confirm an appointment.
My next step is to expand operational testing and prepare the system for practical use. Longer term, I want to explore how this approach could help other small businesses and independent professionals who have limited resources for administrative support.
Log in or sign up for Devpost to join the conversation.