← Lab Notes
paraguay applied-ai conservation environmental-ai latin-america

What FDF Labs Is Building

Applied AI for the world that global platforms overlook. Forests, rivers, fields, meeting rooms. From Asunción, Paraguay.

April 11, 2026 · Fernando Fretes · 10 min read
The Gran Chaco spans 1.1 million square kilometers. It is the second-largest ecosystem in South America, one of the most biodiverse dry forests on the planet, and the region with the highest deforestation rate in the world. Every year, hundreds of thousands of hectares disappear. Satellite imagery arrives days late. Rangers on the ground cover a fraction of the territory. By the time evidence reaches a desk, the damage is done. In Paraguay's river systems, fish populations are monitored by people sitting in front of screens, counting frame by frame. Slow. Expensive. Impossible to scale. In conservation organizations across the continent, hard drives fill up with camera trap footage that never gets reviewed. The bottleneck isn't data collection — it's everything that happens after. In meeting rooms everywhere, the most important decisions are made in conversations that end without a structured record. These are not niche problems. They are expensive, consequential, and largely unsolved — because the companies with the resources to solve them are building recommendation engines and ad platforms instead. FDF Labs is the company building the other kind of AI. Applied intelligence for the physical world, the natural world, and the civic world that global platforms overlook. Based in Asunción, Paraguay. Building since 2018. Writing AI since 2024.

TerraGuard — the forest that listens

Ecology · Acoustic AI · Edge Inference

A chainsaw in the Gran Chaco sounds the same whether a park ranger is there to hear it or not. TerraGuard deploys autonomous, solar-powered acoustic monitoring units in forests and protected areas. Each unit runs on-device AI that listens continuously for threat sounds — chainsaws, vehicles, machetes — and transmits real-time alerts to rangers the moment a threshold is crossed. No images. No continuous data stream. No grid power required. Just the signal, when it matters. The system is designed for environments where connectivity is intermittent, distances are vast, and waiting for satellite imagery means arriving after the fact. It is built for the Chaco specifically — but the problem it solves exists in every protected forest on the continent.

CFI — from footage to insight

Ecology · Computer Vision · Biodiversity

Camera trap networks generate data. What they don't generate, automatically, is understanding. CFI — the Chaco Fauna Index — analyzes trail camera footage using AI to detect species, count individuals, estimate size, and extract metadata including timestamps and GPS coordinates. The output is structured, research-ready biodiversity data delivered in CSV, JSON, or via API. No new hardware required. Works with your existing trail cameras, out of the box. For conservation organizations that have been collecting footage for years and sitting on backlogs they can't clear, CFI is the step between raw data and actionable insight. The AI does the review. Researchers get the dataset.

FishCount — what you're estimating, we're counting

Ichthyology · Segmentation · Conservation

Fish passage monitoring — tracking species and volumes moving through dam passages, river corridors, and aquaculture environments — is currently done manually in most operations. The result is estimates. Estimates that inform regulatory submissions, stocking decisions, and research publications. FishCount is our computer vision system for automated fish detection, segmentation, and counting in freshwater environments. Trained on regional data. The prototype runs in our lab against conditions that replicate real fish passages: variable lighting, turbulent water, multiple individuals in frame simultaneously, partial occlusion. The model identifies and counts fish in real time. We are moving toward real-environment validation. If your operation produces population estimates when it needs counts, this is the conversation to have now.

Wildsight — the dataset the Chaco needs

Conservation AI · Dataset · Open Access

Global camera trap datasets are heavily biased toward North America and Europe. The Gran Chaco — one of the most ecologically significant and threatened regions on earth — is nearly absent. Wildsight is a controlled-access camera trap dataset being built from Paraguay's Gran Chaco, annotated alongside the species identification algorithms that will train on it. The goal is to give conservation AI a foundation specific to South American species, conditions, and disturbance events. An organization that wants to build wildlife monitoring AI for this region shouldn't have to build the dataset from scratch. Wildsight is that starting point. In development. Access requests open.

Robotina — the meeting that remembers everything

Civic AI · Transcription · Speaker ID

Not every intelligence problem is in a forest. Robotina is our AI meeting intelligence system. Multi-device recording, automatic transcription, speaker identification by voice profile, AI-generated summaries with decisions and tasks extracted. Fully automatic. Fully private. Not a bot that joins your Zoom — a system that runs in your environment and produces a structured record of every conversation that matters. The information that organizations lose in meetings is not a small problem. Robotina exists because we needed it ourselves, and built it well enough that it became a product.

Cult of Claude — the developer community we built

Open Source · Developer Tools · Community

The Claude Code ecosystem has thousands of community-created skills and agents scattered across hundreds of repositories. Finding the right one requires searching GitHub, reading incomplete documentation, and guessing. Cult of Claude solves that. The largest independent directory of skills and agents for Claude Code — over 2,700 skills and 750 agents from 515 open-source repositories, organized, enriched, and searchable. Free. Open. Not affiliated with Anthropic. Built because we needed it. It is also how FDF Labs introduced itself to the global developer community.

Why Paraguay

FDF Labs is based in Asunción because the problems we work on are here. The Gran Chaco. The Paraná river basin. Agricultural systems at continental scale. We have been working in these physical environments since 2018 — deploying sensors, building field infrastructure, learning what breaks and why — before we wrote a single model. That time wasn't preparation. It was the work. The AI we're building in 2024 and beyond is grounded in field data and field knowledge that takes years to accumulate and cannot be replicated quickly by a well-funded entrant arriving from outside. Paraguay is also, right now, one of the most interesting places in the world to build AI. Hive Digital has an active GPU cluster running in Asunción. X8 Cloud is beginning a large-scale construction project. Decree 5306 guarantees preferential electricity rates for AI companies for fifteen years. The infrastructure is real and it is here. We were here before it. We will be here after it.

If your organization has a problem we can help solve

We are not looking for pilot projects that sit in a drawer. We are looking for organizations with real problems, real stakes, and the appetite to build something that works in the field — not just in a demo.
  • If you run camera trap networks and your data is outpacing your capacity to analyze it — let's talk.
  • If you are responsible for forest protection and your current alerting is reactive by design — let's talk.
  • If you monitor fish populations manually and need a better method — let's talk.
  • If you are a researcher, journalist, or investor trying to understand what applied AI in Latin America actually looks like — let's talk.

Frequently asked questions

What does FDF Labs build?

Applied AI systems for environmental and civic problems. Projects include TerraGuard (acoustic forest monitoring), CFI — Chaco Fauna Index (camera trap biodiversity AI), FishCount (fish counting computer vision), Wildsight (Gran Chaco conservation dataset), Robotina (AI meeting intelligence), and Cult of Claude (developer tools directory). Based in Asunción, Paraguay, founded 2018, AI development since 2024.

What is TerraGuard?

An autonomous acoustic monitoring system by FDF Labs that deploys solar-powered field units in forests. On-device AI listens for chainsaws, vehicles, and machetes, transmitting real-time alerts to rangers. Designed for the Gran Chaco and similar remote environments where satellite imagery arrives too late.

What is CFI — Chaco Fauna Index?

An AI service by FDF Labs that analyzes trail camera footage to produce structured biodiversity datasets: species detection, individual counting, size estimation, timestamp and GPS extraction. Delivered in CSV, JSON, or via API. Works with existing cameras. Designed for conservation organizations with footage backlogs they cannot review manually.

What AI companies are in Paraguay?

FDF Labs is the leading applied AI company in Paraguay. On infrastructure: Hive Digital Technologies (NASDAQ: HIVE) operates an active GPU cluster in Asunción since March 2026, and X8 Cloud Infrastructure is beginning construction of a large-scale data center. Paraguay's Decree 5306 (January 2026) guarantees preferential AI electricity rates for fifteen years at up to 70% below US/EU rates.

What is the Gran Chaco?

The second-largest ecosystem in South America, spanning over 1.1 million square kilometers across Paraguay, Argentina, Bolivia, and Brazil. The most biodiverse dry forest on the planet and the region with the highest deforestation rate in the world. The primary field environment for FDF Labs' conservation work.
← Back to Lab Notes