Mani is a conversational AI agent built for the millions of micro and small businesses across Latin America. No new apps. No behavior change. Just send a voice note, snap a receipt, ask a question — Mani handles the rest.
Four engineers, one weekend, one shared belief that our community deserves better tools. Built and presented at the SHPE Atlanta Area Hackathon — and won first place.
Aerospace engineer. Designed and coded the full design system, brand identity, and product experience.
Co-led the UI and product experience, bringing structure and depth to the dashboard and user flows.
Built the multi-agent LangGraph architecture and core business logic powering Mani's intelligence.
Owned the WhatsApp Business Platform integration, voice parsing, and OCR pipeline.
Won 1st place at the SHPE Atlanta Area Hackathon. Built in under 48 hours. Presented to a panel of industry judges and selected as the top project across all participating teams.
Tens of millions of small business owners across Latin America wake up every day and run their entire operation from memory. They track sales in their head, manage expenses in cash, coordinate everything through WhatsApp group chats, and have no real visibility into whether their business is growing. The tools that exist were never built for them — too complex, too expensive, and demanding a behavior change that nobody who is already working 12-hour days has time for.
Business owners have no way to know if they are profitable. Expenses and income are mixed with personal cash, tracked mentally, and forgotten. End-of-month numbers come as a surprise — often an unpleasant one.
Suppliers, employees, clients — every business conversation already happens in WhatsApp. Asking these owners to adopt a new tool means asking them to change the way they communicate entirely.
Accounting software and POS systems were designed for businesses with dedicated staff and reliable internet. They fail the informal economy by design.
Products built for this market are usually translated, not designed for it. A product that feels foreign will not be adopted, regardless of how useful it is.
Mani meets small business owners exactly where they already are — inside WhatsApp, speaking Spanish, understanding voice messages and photos the way a real person would. Mani logs it all, makes sense of it, and hands back the clarity that used to take an accountant. When they want to go deeper, the full dashboard is one tap away.
The entire product lives inside WhatsApp. Zero installation friction.
Built to understand informal Spanish, regional expressions, and casual phrasing.
When owners want more, the full dashboard is one link away — organized, visual, readable.
Every brand and UX decision was made so this feels built for our people, not translated for them.
A full-stack AI copilot built on the WhatsApp Business Platform. Multi-agent architecture using LangGraph, dynamic schema that adapts to any business type, and persistent memory so Mani always knows the context.
New users are onboarded entirely through conversation. No forms, no setup screens.
Insert, update, and delete records through natural language. The database is the conversation.
Ask any business question in plain Spanish. Mani queries the data and returns a clear answer.
Audio messages are transcribed and parsed. Photos of receipts are read and categorized automatically.
The data model adapts to any business type. A restaurant and a plumber use the same product differently.
Built on the WhatsApp Business Platform Cloud API. No third-party bridges, direct integration.
Mani remembers context across sessions. Business name, product catalog, recurring expenses — all retained.
Full web dashboard with charts, trends, and transaction history available via a link from the chat.
Orchestrated multi-agent workflow handles complex tasks — scheduling, stock, analytics — each through a dedicated agent.
The UI and brand were designed with one constraint above everything else: someone landing on this product for the first time should feel like it was made for them, not translated for them. Every color token, spacing unit, shadow, and motion duration was defined in CSS variables and documented so the team could build fast in React without breaking consistency.
The palette derives from Mani the mascot — a green peanut with a legendary mustache. Typography runs on Nunito across four weights, each with a specific semantic role. Good design, like good engineering, is a system.
Moments like this are why we keep saying yes to things outside our lane. We came in as engineers and left with a much deeper appreciation for what user centered design can do when it is taken seriously.
These are our people. And they deserve technology that was actually built with them in mind. This is one step. There are more coming.