Rodrigo de Miguel

TalkFlux started as an internal tool to solve a need I had myself: access to the best models from different providers without maintaining multiple subscriptions, using the APIs and paying only for actual usage.

The first version was very simple: conversations, projects, and model switching during a chat. Seeing that the problem was quite common, I decided to turn it into a product.

That involved much more than the chat: authentication, access control, Free/Flex/Own Key plans, usage tracking, a prepaid wallet, BYOK, and billing with Stripe.

The product is still small and evolving, but it is already in production and open to new users. The next stage is expanding its capabilities without losing the simplicity that motivated it.

With everything I learned while building ADOPTA, I started SietchStack as a way to take it a step further: a multi-tenant platform for launching multiple niche directories from a single codebase, with data ingestion, LLM-based enrichment, and an SEO-first approach.

Every product is defined by configuration — entities, routes, sitemap, language, design and components — without touching the application core. That let me iterate and launch new directories focusing on the infrastructure rather than on a single product.

I shelved it when AI Overviews changed the playing field and the SEO thesis I started with stopped holding up. The foundation is built, and if the landscape becomes interesting again I would not rule out picking it back up.

I built a natural dog and cat food brand from the ground up and took it from concept to nationwide sales. Product, brand identity, e-commerce, and operations: everything had to fit together. It was an exercise in pure bootstrapping, where every decision counted and every euro had to be put to good use.

I had to do it all. Design pet food and treats, talk to and negotiate with suppliers, work on packaging, set up and customize Shopify, coordinate the warehouse and shipments, assist customers, and close sales. When I needed to learn something new, I learned it. When something needed to be solved, I had to do it.

Technology was there to multiply resources. I built automations, email marketing workflows, and customer acquisition and retention processes; I connected the e-commerce platform with the rest of the ADOPTA ecosystem and experimented with advertising, funnels, and SEO. There wasn’t a huge team behind it. I had to make sure all the pieces worked together.

And there was a reason for doing it that way. The proceeds from food sales helped fund ADOPTA, the free platform for animal shelters and rescue organizations. Every purchase had a second purpose: in addition to feeding a dog or cat, it helped sustain a platform created to help more animals find homes.

ADOPTA was born from something much simpler than a business idea: seeing a problem up close. While I was studying and working, I volunteered at an animal shelter in my town. There, I discovered a reality that was hard to ignore: shelters with limited resources trying to care for animals while also ensuring that the right people could find them.

I started Adopta un animal near the end of my college years. I developed it during my master"s program while juggling my job, gradually turning an initial idea into a platform designed to address a real problem: making it easier to post, find, and manage animals up for adoption.

Over time, it became the brand ADOPTA. I built the product from start to finish: architecture, backend, frontend, admin panels, and the infrastructure needed to grow organic visibility. On one hand, animal shelters could manage their animals; on the other, adopters had a simple experience to discover dogs and cats, filter results, and learn each animal"s story.

Over time, ADOPTA ended up being much more than just the first product I built. It became the starting point for a family of projects centered around animals, data, SEO, and automation. That's where I learned that building a product means first understanding the problem and then finding the right technology to solve it.

Finding an animal shelter in Spain should be easy. It wasn't. The information was scattered across directories, websites, maps, files, and local pages, often with different names and incomplete details. I wanted to bring it all together in one place and make finding an animal shelter as simple as looking at a map and seeing what was nearby.

I built a data ingestion engine in Node.js capable of collecting data from a wide variety of sources—APIs, scrapers, files, and open data—and converting it into consistent information. The challenge was to group together entries belonging to the same location, remove duplicates, and normalize unstructured data: entity resolution, text similarity, blocking, and data cleansing all working together to support a simple user experience.

The result was a directory with more than 1,700 shelters and animal rescue organizations—the largest in Spain—displayed on a map and searchable by location and filters. I later expanded the idea to also include veterinarians and other pet services, creating custom listings and programmatically generated pages to give each business visibility.

Pet Friendly was born out of a very specific need and ended up becoming one of the projects that taught me the most about data, SEO, and reusable systems. What began as a map of animal shelters would end up laying part of the foundation for what I would later build with SietchStack.

When GPT-3 and GPT-3.5 first started appearing, back in 2022, my curiosity was piqued: How far could generative AI go if I connected it to structured data and programmatic SEO? Within ADOPTA, I set up a small lab to find out. The starting point was Google’s People Also Ask; the goal was to automatically generate useful content about virtually any dog or cat breed.

I compiled information, images, and data on 336 dog breeds and 94 cat breeds, designed the generation system, and ended up publishing more than 60,000 articles on characteristics, care, and nutrition. All built on WordPress, with LLMs processing structured data and an architecture designed to produce content at a scale that would have been impossible to write by hand.

It was probably one of my most fun experiments. It was also one of the ones that best taught me where tricks end and reality begins. The project lasted until Google deactivated the domain during a Core Update. Looking back on it from 2026, with generative AI having become an everyday tool, it seems almost prehistoric. But that’s precisely why I’m keeping it: because it was one of the first times I truly saw how data + code + AI + distribution could become a system capable of producing something on a completely different scale.

For a while, I helped some local businesses take their first steps into the digital world. I worked on their Google Business Profile and their visibility in local searches, turning businesses that barely had an online presence into places that were easier to find.

It was a small phase, almost parallel to the rest of my projects, but it helped me get back to something very basic: technology only matters when it brings someone closer to a real business. A better-crafted profile, a well-optimized search result, and a customer who finds what they need can be enough to change the story.