Roles · Engineering
Founding AI Engineer
Latin America will spend this decade deciding whether it builds frontier AI capability or rents it. The region holds the energy, the minerals, the data, and the talent of 660 million people, and almost none of the compute, the models, or the governance leverage. Futuros exists to change that.
Futuros is the evidence platform for that argument: a live, cited, trilingual observatory of 25 countries built to put verifiable numbers in front of heads of state, CELAC, the IDB, CAF, ECLAC, and the private sector. Every claim on the platform resolves to a source, and the platform is in production today.
Platform in early access for partners and funders. We will walk you through it live during the process.
The role
The founding AI engineer owns the reasoning layer: everything between a user's question and a grounded, cited answer. Retrieval, grounding verification, evaluation, agent systems, and the path to sovereign inference. You work directly with the founding CTO, and what you ship goes in front of ministers and multilaterals.
What you'll do
- Build the retrieval stack. Hybrid lexical and semantic indexing over a corpus of thousands of documents and datasets, reranking, and query understanding that works in Spanish, English, and Portuguese.
- Own grounding verification. Every generated answer must resolve to cited evidence. Build the gates that catch an ungrounded claim before a user ever sees it.
- Build the evaluation harnesses. Faithfulness, freshness, confidence calibration, and regression suites that gate every deploy. If quality drops, the build fails.
- Engineer the agent systems. Orchestration, tool design, and observability for the AI-agent workforce that maintains and extends the platform.
- Prototype the sovereign-inference path. Fine-tuning experiments, serving benchmarks, and cost and latency profiles for running regional models on regional infrastructure.
- Run inference in production. Provider failover, caching, batching, token budgets, and the failure modes of streaming systems under real load.
What we're looking for
- AI-native. Agents, models, and evals are how you build, not a feature you bolt on. You have shipped LLM systems to production and can reason about their costs and failure modes.
- Retrieval depth. You have built search or RAG systems that real users depended on, and you can explain why they worked.
- Engineering fundamentals. Strong TypeScript or Python; you write code that other people can run, test, and extend.
- Evidence discipline. You would rather return a smaller, verified answer than a fluent, unverifiable one.
- Founding temperament. You are comfortable owning ambiguous problems with no one above you to escalate to.
Nice to have
- Working Spanish. Strongly preferred; our corpus and our users lead with it.
- Roots or deep working ties in Latin America.
- Multilingual NLP or cross-lingual retrieval experience.
- Fine-tuning and model-serving experience beyond API calls.
- An open-source track record.
How this works
This is a founding conversation. We are open on compensation, equity, and scope for the right fit, and we come to the table with strong institutional backing. We will be transparent about terms from the first call and expect the same directness back.
How to apply
Send a short note on your interest in the role, an example of what you've built that's aligned with the role, and why this is the dream role for you.
marcus@ladp.io