Regional evaluation sets
Task sets built from authorized source material in es-CO, es-MX, es-AR and other variants, for the workflows where the country changes the right answer: tax, clinical pathways, road freight, banking, public procurement.
For data and AI companies
Your Latin American delivery partner for expert data.
Savia sources, verifies and manages Spanish-speaking domain experts across Latin America, and delivers evaluation, annotation and workflow-capture work under your rubric, with provenance on every item. White-label or co-branded, on your platform or ours.
What you can hand to us
Five things a partner outsources. One accountable team.
Expert sourcing and verification
We recruit practitioners in the countries you need, interview them in Spanish, and check licences against the national registry where one exists. In Colombia: ReTHUS for health professionals, COPNIA for engineers, the Registro Nacional de Abogados for lawyers, the Junta Central de Contadores for accountants. In El Salvador: the Consejo Superior de Salud Pública for health professionals, the CVPCPA for accountants, the Corte Suprema de Justicia register for lawyers. You receive a vetted roster, not a spreadsheet of self-declared titles.
Managed evaluation and annotation
Your rubric, our reviewers, our QA layer. RLHF preference ranking and SFT examples, pairwise preference, rubric scoring, red-teaming in regional Spanish, transcript grading, structured annotation. Every item records who did it and whether it was observed, derived or confirmed by a person.
Workflow capture
WorkTraces: how a unit of real work is actually done, step by step, with the exceptions and judgment calls, recorded from practitioners and confirmed by a named reviewer. The training asset that generic data cannot produce.
Field data collection
Egocentric video and real-world task capture for AI and robotics teams. We recruit, train and manage the collaborators who record, with household consent under local law and daily quality review against your acceptance criteria.
How an engagement runs
Scoped like a vendor. Run like a team of yours.
Scope
A statement of work with the task definition, rubric, volume, turnaround and acceptance criteria. We say what we can staff and by when. We do not accept work we cannot deliver.
Staff
Experts matched by country, discipline and credential. Contractor agreements under Colombian or Salvadoran law, payment in US dollars, data-protection consent from every person before work starts.
Deliver
On your platform or ours. QA sampling on every batch, a named lead on your side and on ours, a weekly quality report you can forward as it is.
Audit
Per-item provenance, reviewer identity, consent records, deletion on request. Nothing model-inferred is ever delivered as confirmed.
Who this is for
If your customers need Latin America and your bench does not have it.
Expert marketplaces and data vendors
Add verified Latin American coverage without opening an office. We run as your regional pod, under your brand or ours.
AI labs and evaluation teams
Regional Spanish evaluations built and scored by practitioners who work in that country, with the error analysis that explains a failure.
Enterprise AI teams expanding south
Test an agent against Colombian, Mexican or Argentine reality before it reaches a customer there.
Where we are today
Early, and precise about it.
Savia is a young company working across Latin America, with a working platform, a benchmark program in road freight and clinical review, and an expert pool that is growing through this site. We do not publish headcounts or client names we do not have. Ask us for the current numbers and you will get the exact ones.
Questions
What partners ask first.
Do you do RLHF and preference data in Spanish?
Yes. Pairwise preference ranking, rubric scoring, SFT example writing and red-teaming in regional Spanish variants such as es-CO, es-MX and es-AR, done by practitioners of the discipline, with reviewer identity on every item.
Can you run field data collection, such as egocentric video?
Yes, per project. We recruit, train and manage the collaborators who record, with household consent under local law and daily quality review. The field collection page explains how a project runs.
Which countries do you cover?
We work across Latin America. Professionals register from every country, and we staff each country project by project when we can verify the people there.
What is the minimum engagement?
A pilot of 100 expert hours or USD 5,000. Hourly bands are published on the rate card.
Can you work under our brand?
Yes. White-label or co-branded, on your platform or ours, with no platform fee.
AI data services
Services for AI and data teams
Spanish AI evaluation for Latin American markets
Regional Spanish AI evaluation for es-CO, es-MX, es-AR and other scoped variants: rubric scoring, preference ranking and red-teaming.
Explore the serviceExpert annotation for healthcare, law and finance
Healthcare, legal and finance annotation in Latin America with project-specific practitioner verification, source rights and item-level provenance.
Explore the serviceEgocentric video collection in Latin America
Project-based egocentric video collection in Latin America: recruitment, local-language onboarding, household consent and daily quality review.
Explore the serviceRegional benchmarks built from verifiable sources
Latin American AI benchmarks built from authorized sources, explicit derivations and regional task definitions, with model-inferred content clearly labeled.
Explore the serviceSOW / ROSTER / QA / AUDIT
Talk to us about a pilot
Tell us the task, the country and the volume.
We reply with what we can staff, how we verify it, and what it costs. Usually within two working days.