Data. Strategy. Intelligence.

AI that works in the real world.

We build production-grade AI systems across three domains: intelligent agents, decision-support platforms, and immersive digital experiences — deployed across four continents.

Start a project Explore our work

52+

Active AI Deployments

10+

Products Delivered

99.9%

Accuracy Rate

Who we are

Lean team.
Production scale.

Apis-IA is an AI product studio focused on intelligent, data-driven software across three strategic pillars. Operating as a lean, high-precision team, we combine deep expertise in large language model integration, retrieval-augmented generation (RAG), full-stack engineering, and human-centered product design to deliver production-grade solutions for public and private sector clients across four continents.

Our portfolio spans 52+ active AI deployments and projects with tourism boards, hotel groups, national parks, airlines, educational institutions, environmental agencies, and enterprise intelligence platforms — each built for scalability, domain awareness, and architectural soundness.

AI & LLM Integration Data Engineering PWA & Mobile Full-Stack Delivery AI Agent Design Cloud & DevOps

The Founder

Three decades of engineering.

Fernando Lopez

Founder & Technical Lead

Fernando brings 30+ years of continuous software engineering — from the early commercial internet to today's AI-driven product generation. His background combines rare depth across the full arc of web development, with the commercial instinct of a founder and hands-on engineering experience across two continents.

ColdFusion Java C# / .NET MS SQL Server Front-end Back-end E-Commerce AI / LLM

Early 1990s

Started in technology in Brazil, building foundational software engineering skills during the rapid expansion of commercial computing.

1996 — US

Led development of foodtrade.com, one of the first vertical sites dedicated to international food trade — predating mainstream e-commerce.

1996–2014

18 years delivering projects across the full evolution of web development: ColdFusion, Java, C#/.NET, MS SQL Server, modern frontend and backend. Built and operated an independent e-commerce platform.

Today

Recognized the transformative potential of LLMs and generative AI. Founded Apis-IA with deliberate focus on AI-driven engineering — leading technical strategy and product delivery across multiple sectors and geographies.

Pillar 01 — Agents

Domain-aware agents.
Zero hallucination.

Conversational AI systems configured with domain-specific knowledge, terminology, and retrieval logic — built for tourism, hospitality, public services, and consumer engagement.

VisitAssist™

Multi-Tenant Conversational AI Platform

The flagship platform behind 52+ active deployments across tourism boards, national parks, luxury hotels, airlines, and public libraries on four continents. Built on a semantic RAG pipeline, VisitAssist delivers instant, factually-grounded answers from curated knowledge bases — with zero hallucination on verified data.

Semantic RAG pipeline
Multi-tenant scalable architecture
Web embed, PWA & deep link deployment
Per-instance knowledge base & brand config

Deployment variants

VisitAssist — Tourism GuestAssist — Hospitality BiblioTecno — STEM Education Global Pilots — 4 Continents

Toonli

AI Creative Media Platform

AI-powered platform for image and video generation via conversational prompts, deployed as a VisitAssist agent implementation. Full commercial feature set with multilingual delivery and an integrated affiliate commerce engine.

Multimodal image & video generation
Multilingual delivery (EN/PT/ES/ZH)
Affiliate management & commission tracking
FFmpeg media processing pipeline

Food Trails

Culinary AI Concierge

Domain-aware AI agent for culinary discovery. Understands user preferences, dietary requirements, budget, cultural interests, and location context to generate personalized gastronomic itineraries through a multi-step planning flow with typed domain models and normalized JSON responses.

Multi-step itinerary planning
Dietary & preference awareness
Typed domain models & JSON normalization
Location & cultural context

Pillar 02 — Intelligence

RAG platforms.
Actionable answers.

Platforms that transform complex, specialized data into AI-queryable, searchable, decision-support products for environmental, commercial, and enterprise audiences.

InsightAssist™

Private BI & RAG Intelligence

Hybrid search architecture combining semantic vector retrieval with keyword metadata filtering — higher relevance and precision across complex, multi-source datasets. AI-assisted query refinement before retrieval. Structured outputs ready for BI charts and executive reports, not just conversational answers.

Hybrid semantic + keyword search
AI-assisted query refinement
BI-ready structured outputs
Modular frontend / API / retrieval architecture

CarbonBiz / CarbonIntel

Carbon Market Intelligence Platform

Specialized information platform for carbon market news and intelligence — translating complex policy, regulatory, and environmental data into a usable digital product. Domain-oriented information architecture with a structured filter taxonomy driven entirely by configuration, not hardcoding.

Config-driven filter taxonomy (region, sector, credit type, policy)
Multilingual EN/PT/ES/ZH via i18next
Persistent filter preferences across sessions
Offline PWA via service worker

Pillar 03 — Complexity Studio

Immersive products.
Difficult subjects.

Educational and experiential products that combine AI, human-centered design, and architectural discipline — turning complex domains into explorable, interactive systems.

CrisisLab

AI-Powered Crisis Simulation

Deterministic rules engine paired with a structured LLM narration layer. Clear architectural separation between game logic (state machine, branching) and AI narration. OpenAI integration with TypeScript typed contracts and JSON Schema-based structured outputs — production-validated responses throughout.

Deterministic rules engine + LLM narration layer
TypeScript typed AI contracts
JSON Schema validated outputs
Scenario branching & real-time consequences

Bookworld

AI Literary Content Platform

RAG pipeline built on Next.js 15 for long-form text ingestion and structured entity graph extraction — characters, places, themes, and timeline. Human-in-the-loop editorial workflow with distinct stages: intake, extraction, synthesis, enrichment, review, and publication.

Semantic RAG for long-form literary content
Entity graph extraction (characters, places, themes)
Human-in-the-loop editorial workflow
Structured knowledge synthesis pipeline

How we build

Retrieve. Reason. Deliver.

Retrieve

Semantic vector search across domain-specific, structured knowledge bases — ensuring the right information is always surfaced before the model reasons over it.

— Semantic vector retrieval
— Hybrid keyword + semantic search
— Metadata filtering
— AI-assisted query refinement

→

Reason

Frontier models operate over retrieved, grounded context — with typed contracts and JSON schema validation ensuring structured, production-safe outputs at every step.

— Frontier LLM integration
— Structured output contracts
— JSON Schema validation
— Deterministic logic separation

→

Deliver

Actionable results through the right channel — conversational interfaces, BI dashboards, PWAs, or API integrations — with multilingual support and offline capability where needed.

— Multi-channel deployment
— BI-ready structured outputs
— PWA & offline-first
— Multilingual (EN/PT/ES/ZH)

Engineering Principles

How we think.
Why it holds up.

Domain-Focused AI

We don't apply generic models to specialized problems. Every deployment is configured with domain-specific knowledge, terminology, and retrieval logic tailored to its context.

RAG

RAG as Standard

Retrieval-augmented generation is our default for any system requiring factual precision. Hallucination is eliminated by grounding every response in indexed, verified source material.

→

Production-First Delivery

Every project includes build validation, environment configuration, deployment documentation, and a clear path from MVP to enterprise hardening — not just a working prototype.

{}

Structured Outputs

Every AI-assisted feature uses typed contracts, JSON schemas, and validated response parsing — reducing operational risk and enabling deterministic rendering across surfaces.

⇄

Separation of Concerns

Deterministic logic — state machines, rules engines, validation — is always separated from probabilistic AI layers. As evidenced in CrisisLab and Bookworld, the two never entangle.

∞

Scalable Platform Thinking

We build platforms, not one-off solutions. The 52+ VisitAssist deployments running from a single architecture are the clearest proof of this principle in production.

Full-stack. Production-grade.

Beyond our flagship platforms, Apis-IA has independently delivered complete digital products across education, public safety, trade intelligence, and creative tooling.

Full-Stack Product

LITERAH

Python / Flask · Supabase / PostgreSQL

Multi-profile educational reading platform supporting students, teachers, principals, and administrators — each with differentiated dashboards and role-based access. Includes quiz system with question banks, reading progress telemetry, and SQL migrations with RPC functions.

Full-Stack Product

Vayya

React · Supabase · Real-time

Personal safety app (MVP) with real-time security sessions, a four-state SOS machine (triggered → sent → acknowledged → resolved), encrypted evidence vault with biometric/PIN unlock, and a discreet appearance mode with rapid exit and redacted notifications.

Data Intelligence

Trade Intelligence Explorer

React · TypeScript · Node / Express

Full-stack explorer combining UN Comtrade / ComexStat international trade data with trade fair exhibitor datasets in a unified interface. Defensive JSON parsing, deterministic deduplication by compound key, faceted filters, full-text search, and CSV export.

PWA / Frontend

CineTrip

React 18 · TypeScript · PWA

Installable PWA for discovering real filming locations. Component-based architecture with typed domain models for locations and routes, client-side search and filtering, custom service worker offline cache, and static hosting on Vercel.