The Philosophy & Vision
Hungry is built on a raw, uncompromising philosophy: hunger. Two technical teenagers (17 and 18) from Spain are building the kind of software they wish existed in their families' businesses — real, sovereign, auditable systems that actually move money and paperwork, not another pretty chatbot wrapper.
Our name, Hungry, mirrors the desperate operational hunger of local SMEs (distributors, dealerships, real estate agencies) and resource-constrained town halls that are starving for efficiency. They are drowning in paper, WhatsApp chaos, and legacy ERPs because "enterprise" IT is too expensive, slow, and over-engineered for their reality. Hungry gives every local worker an elite executive assistant on their phone through a simple WhatsApp voice note — compressing hours of manual work into seconds.
The Problem
We target the backbone of Spain's real economy: family-owned B2B SMEs and mid-sized municipalities (10k–80k population) suffering brutal administrative bottlenecks every day.
HORECA Food & Beverage Distributors Bar owners place orders at 2:00 AM through noisy, chaotic WhatsApp voice notes. Admin staff waste 4–6 hours every Monday morning listening and manually entering these orders into legacy ERPs (Sage, SAP), introducing ~10% error rates that cause expensive delivery rejections and lost revenue.
Automotive Finance & Luxury Real Estate High-ticket leads go cold during evenings and weekends because salespeople are offline — dropping conversion rates from ~25% to under 5%. Collecting client documents (IDs, pay slips, proof of address) over chat is slow, manual, and error-prone.
Immigration & Public Licensing (Extranjería, municipal permits) Appointments are blocked by automated bot scalpers hoarding slots and reselling them for 50–200 €. Flows that require Cl@ve PIN SMS authentication break most automation attempts, forcing human staff to click through every step manually.
How It Works
Hungry is a voice-first agentic infrastructure that turns WhatsApp audio into secure, API-driven or browser-driven real-world actions.
1. Ingestion via WhatsApp The user sends a voice note: "Book a luxury car test drive for Carlos this Saturday, check his documents, and print the confirmation." Hungry ingests raw audio including background noise and informal speech.
2. Cognitive Loop with Gemini Gemini 2.5 Flash transcribes and parses the audio, extracting structured parameters (customer name, DNI, vehicle model, time window, discount rules) using strict JSON schema enforcement at temperature 0 — delivering sub-second extraction with 0% formatting errors.
3. Human-in-the-Loop Security Shield (HITL) Before any external tool fires, Hungry sends a secure 2-step approval link to the salesperson's phone. They review the parsed action, edit any field if needed, and approve or reject. No transaction, appointment, or document is executed without explicit human consent.
4. Decoupled MCP / Workflow Execution Once approved, Hungry triggers decoupled Model Context Protocol (MCP) servers or n8n webhooks to update Salesforce/CRM, register the order in the ERP, generate a signed PDF, and physically trigger a local office printer via a lightweight on-prem connector.
5. Autonomous Browser Worker for "no-API" Portals For government or legacy portals without APIs (like Extranjería), a dedicated worker runs agent-browser via residential proxies to navigate through Cloudflare, CAPTCHAs, and multi-step forms. When a Cl@ve PIN SMS or any 2FA is required, Hungry pauses, asks the user for the code over WhatsApp, and securely injects it into the session — completing the process without breaking compliance.
Why Gemini?
Hungry achieves near-zero operating costs and extremely high gross margins by leveraging Gemini's state-of-the-art capabilities.
- JSON Schema Enforcement (Gemini 2.5 Flash): Strict Pydantic schemas at temperature 0 guarantee 0% formatting errors during sub-second extraction, keeping our FastAPI backend stable and per-call costs under €0.00025.
- 1 Million Token Context Window (Gemini 2.5 Pro): For municipal workflows, we inject entire urban planning law PDFs — hundreds of pages — directly into the model's context, enabling full legal compliance validation without the recall issues of traditional RAG pipelines.
- Native Multimodal Vision: Gemini inspects photos of IDs and pay slips sent via WhatsApp, instantly flagging documents that are blurry, expired, incomplete, or inconsistent before a human ever touches the file.
Business Model
Ultra-affordable, transactional SaaS designed for mass adoption among SMEs.
| Tier | Price |
|---|---|
| Setup Fee (one-time) | €299 |
| Base Subscription | €99 / month |
| Usage Fee | €0.05 / transaction |
| Municipal Annual Pilot | €14,500 / year |
Unit Economics: A typical SME running 1,000 transactions/month pays ~€149. Our underlying Gemini + infra costs for those 1,000 calls: ~€0.25. Gross margin: ~99.8%.
Municipal pricing fits the Spanish contrato menor ceiling — direct contracting without public tender — letting us prove value inside one or two departments before scaling to the full administration.
The Ambition
Our ambition is to build the sovereign, voice-first agentic operating system for Europe's real economy. While big tech races to build search chatbots and generic web agents, Hungry focuses on the boring, painful workflows that actually move the economy: orders, permits, financing, and service tickets.
Spain is our proving ground. If we can make autonomous agents work safely under GDPR, LOPD, and the Spanish Esquema Nacional de Seguridad (ENS), integrating with the reality of Cl@ve, legacy ERPs, and public portals — we prove that a secure, localized, human-verified agentic network can scale across Southern Europe and beyond.
Winning the Build with Gemini XPRIZE would give us the capital and credibility to hire top-tier engineers, certify Hungry under ENS (medium level), expand to thousands of SMEs and dozens of municipalities, and show the world that age is just a number when relentless hunger meets state-of-the-art AI.
Log in or sign up for Devpost to join the conversation.