We are an
AI engineering studio
and Venture Builder
80 people
65% software engineers.
35% project managers, strategists, designers, and support staff.
Under 10% turnover rate among developers.
Remote Company, Small Teams, Full Ownership
Remote and asynchronous from day one. We organize engineering in small, self-contained units, with a named owner accountable for each project end to end. AI takes on more of the execution, but judgment, architecture, security, and results stay with people.
About
We take on complex technical challenges that have to hold up in production.
We build enterprise AI for Fortune 500 companies, public institutions, and scale ups: systems that survive real data, real users, and real compliance. 30+ of them are running today across the US, LATAM, Europe and the Gulf, several under strict data residency and compliance regimes.
Our Venture Builder applies the same engineering to our own AI-native products in regulated, high-stakes markets. The ventures keep the services honest, and the services keep the ventures grounded.
We are bootstrapped and financially independent. That lets us invest in people, research, and products on a longer horizon than a client contract.
Engineering
We are a Select Services Partner in the Claude Partner Network.
We work on
Search and information retrieval · Agentic systems · Evals and observability · Event-driven architectures · Data engineering · Platform modernization.
We build it with
TypeScript · Python · Java · Kotlin · React · Node · Kubernetes · AWS · GCP · Kafka · PostgreSQL · Elasticsearch and OpenSearch · Vector databases · Terraform · OpenTelemetry · MCP.
Language
English is our working language.
Spanish is our native language.
Our team is internationally distributed.
Experience
- 2026NOW
We now operate as a fully AI-first boutique, in the work we do for clients and in how we run the company itself. TAM OS, our internal platform, is both our operating core and the technological foundation we bring into client projects. Our Venture Builder bets are starting to take off: nau! in mortgage switching, Zaina in the Middle East, and more coming. We are building relationships with frontier labs and top partners, including our Select Services Partner status in the Claude Partner Network.
- 2024-2026CONSOLIDATION
The period when the shift became structural rather than directional: more than half of services revenue was coming from AI. We launched the Venture Builder to build and own products in market opportunities we could see from client work, and to push the technology further than a client engagement usually allows.
- 2020THE BET
We had been waiting more than a decade for a shift like this one. We decided to specialize the whole company rather than add a practice on the side: an independent AI research unit, and our first product unit, which years later became the Venture Builder. The focus was clear from the start: search, information retrieval, and enterprise AI in complex environments. AI Findr was the first major result.
- 2015-2018US EXPANSION
In 2015 we went to the US and eBay's innovation team bet on our ideas. They became our first client, and from there the work itself did the selling: senior technical leaders vouched for us, and their teams passed us on to other engineering teams. That snowball is how a small team from the Canary Islands ended up in deep, long-term relationships with Fortune 500 companies and institutions like Harvard and PayPal. We grew without a sales team.
- 2013ORIGINS
Jaime López and Javi Toledo started a small software studio in the Canary Islands: The Agile Monkeys. The idea was clear from day one: solve real, complex problems that few teams were taking on, stay independent, and build something with a positive impact on their lives and on the place around them. They gathered a small group of technical, creative people and got to work.
What we publish, and what we build
Open-source engineering, shared learning, and AI-native ventures
Facility - AI SDLC - open source
Our open-source platform for running AI coding agents inside a reviewable delivery process. Agents propose and implement; humans approve the plan, validate the preview, and merge. Agents never approve or merge anything. It runs our own delivery today, and we published it early on purpose: a platform for governing AI-assisted delivery is not something to design in private for a year. Apache 2.0, v0.3, rough edges included.
The Lab
What we find while building: retrieval regressions that start in the ingestion pipeline, 64 controlled runs benchmarking document parsing across five providers, a legacy modernization that went from 229 vulnerabilities to zero. Numbers and methodology attached, so you can disagree with them.
Booster
Our open-source TypeScript framework for event-driven backends: CQRS and event sourcing with the infrastructure inferred from the code. Apache 2.0, in production across startups and large enterprises, 3,000+ commits and still shipping. We started it years before the current wave, and it is where a lot of our event-driven thinking comes from.
Venture Builder
AI Findr
Enterprise AI search that understands intent and answers from complex, messy corpora. Deployed across shopping centers, banking and healthcare in LATAM and the Gulf, several under strict data residency requirements.
nau!
Watches a mortgage against the live market 24/7 and tells the user when switching would actually pay off, including when it would not. It reads the mortgage from a photo of a bill, no bank credentials involved, and models the real cost rather than the headline rate: fees, bonuses and the product bundles that quietly change what a mortgage costs. Live in Spain with real customers.
Zaina
A conversational AI for aesthetic medicine in the Middle East, in English and Arabic. The interesting constraints are the ones that say no: grounded in recognised medical sources, reviewed by a medical liaison, no diagnosis, conversations never tied to an identity, and answers kept independent from the clinics the product connects to.