Brand logo
HomeProjectsCase Studies
ContactLet's Chat
Let’s build something real

Have a project idea or need a developer?

I’m open to internships, part-time roles, and small product builds. Send a short brief and I’ll reply with next steps.

Book a CallSend a MessageSee projects

© 2026 Shabnam Beiraghian. Built with Next.js + Tailwind.

ProjectsResumeContactGitHubLinkedIn
Projects
Dew logoWeb · 2025

Dew

Embeddable AI Front-Desk Assistant

Next.jsReactPythonGoogle Gemini
Visit site
ai-client-seven.vercel.app/
Dew

gallery

1/3

Rationale

Context

Most small business websites leave visitors hunting for basic information — hours, services, how to book. Dew solves this by providing an always-on AI assistant that holds real conversations, without requiring a dedicated support team or technical expertise to set up.

Outcome

The result is a lightweight chat widget — powered by Google Gemini, built on Next.js and FastAPI — that any site owner can embed with a single script tag. Business context is loaded from a plain-text knowledge file, making onboarding fast and the system easy to maintain or extend.

Approach

The project needed to work on any website platform with minimal integration effort, support multiple businesses from a single system, and avoid costly infrastructure like model training or vector databases. Every design decision was shaped by the constraint of keeping deployment as simple as possible.

Context

Most small business websites leave visitors hunting for basic information — hours, services, how to book. Dew solves this by providing an always-on AI assistant that holds real conversations, without requiring a dedicated support team or technical expertise to set up.

What it does

  • Dew is a lightweight, embeddable AI assistant designed to give small businesses and portfolio sites an always-on conversational front desk. Rather than requiring complex infrastructure or model fine-tuning, Dew uses structured plain-text knowledge files as system prompt context, allowing any business to be onboarded in minutes.

Approach

The project needed to work on any website platform with minimal integration effort, support multiple businesses from a single system, and avoid costly infrastructure like model training or vector databases. Every design decision was shaped by the constraint of keeping deployment as simple as possible.

Outcome

The result is a lightweight chat widget — powered by Google Gemini, built on Next.js and FastAPI — that any site owner can embed with a single script tag. Business context is loaded from a plain-text knowledge file, making onboarding fast and the system easy to maintain or extend.