Type: Full stack AI SaaS product, built and run solo Deliverable: Live production platform at buyerloop.app Role: Founder, designer, and engineer.

Stack / integrations: Next.js · TypeScript · Tailwind CSS · Firebase · Google Gemini 2.5 Flash · ElevenLabs · Google Calendar API · Resend · Paddle · Vercel

Every business with a website loses enquiries the same way. Someone visits at 11pm, has a question, and no one answers. By morning they have already contacted a competitor. Businesses do not lose enquiries because their leads are bad. They lose them because they were asleep.

The usual fix, a basic chatbot, makes it worse. Generic bots do not know the business, invent answers, and frustrate the exact serious buyer you most wanted to keep. What a business actually needs is not a chatbot. It is an AI that knows their business, talks to their customers like a real team member, works out who is serious, and hands the owner a lead that is ready to act on.

I built BuyerLoop to be that: an AI team any business can put on its website in minutes.

I designed and developed the entire platform from scratch. The marketing site, the dashboard, the embeddable widget, the AI layer, and every backend integration. It is a real multi tenant SaaS with authentication, billing, and per business data isolation, not a prototype. A trainable AI agent, grounded in the business. Each business creates its own AI agent and trains it on its real listings, services, fees, FAQs, and rules through a tabbed editor. They pick an industry, real estate, clinic, therapy and coaching, contractor, or gym, and the agent adapts its language, the questions it asks, and the fields it collects to fit that vertical. Every answer is grounded in the business's own knowledge base, powered by Google Gemini, so it never invents prices, availability, or facts.

Live chat and real time voice, in one thread. Visitors can type or talk. The chat qualifies people conversationally, no forms, pulling out budget, timeline, financing, and intent as the conversation flows. At any point a visitor can switch to a live voice call powered by ElevenLabs, and the agent carries the full context of the chat into that call. It answers spoken questions from the same knowledge base and captures the lead by voice.

New qualified leads land in the dashboard and reach the owner's inbox by email the moment they are captured, through a Resend integration I built into the Next.js backend. From the lead, the owner can send a Gemini drafted, personalised reply by email, and book a viewing straight into Google Calendar through the Calendar API, which sends the buyer a real invite.

A four agent loop. BuyerLoop runs as four agents: Conversation, Qualification, Follow up, and Recovery. With auto follow up enabled, leads that go quiet are re engaged automatically on a schedule, and stalled conversations are recovered, with no manual chasing. Care sensitive industries like clinics and therapy get stricter, safety first prompting so a follow up is never intrusive.

A reports page rolls up every conversation, qualified lead, follow up, and viewing, with a summary written by Gemini in plain business language and a lead activity chart, so the owner always knows what their AI team did.

Secure, embeddable, and domain locked. The whole thing installs with one line of embed code. The widget is locked to the business's own domains with origin bound session tokens, so the snippet only runs on their authorized sites and cannot be copied elsewhere. Dashboard routes are protected by verified Firebase ID tokens, with strict per tenant isolation so no business can ever reach another's data.

I built and run BuyerLoop solo, with AI as the rest of the team, while making every architecture and product decision myself.

Much of the hard engineering is invisible. Making the voice agent inherit the chat context so a conversation stays continuous across the channel switch. Merging chat and voice into one lead.

The go to market was built the same way. Before I reach out to a prospect, I generate a live demo of BuyerLoop running on their own site content, colours, and layout, so they see their own AI agent working before they ever pay. That has been the single strongest thing for turning a cold contact into a paying customer.

BuyerLoop is live in production at buyerloop.app, serving several industries from a single codebase, with subscription billing and a real onboarding flow. It takes a business about five minutes to create an agent, train it, and go live on their site.

It is the product version of a simple belief. Soon every business will have an AI that talks to its customers, sells for them, and handles support, the same way they all have a website, a phone number, and an email today. BuyerLoop is where a business gets theirs.