assistants
AI assistants & copilots
Domain assistants and copilots grounded in your data and tools — accurate, evaluated, with citations.
ai engineering · consulting · systems
Real AI systems, from strategy to production.
relatent is an independent AI studio — classical ML, LLMs, agents and fine-tuning. We help you find where AI actually pays off, then design, build and ship the system and hand your team the keys.
specialty · document workflows
[scope]goal · data · metrics defined
[data]clean · label · split
[model]fine-tune · 3 candidates✓ 0.94
[eval]win-rate 0.94 · in policy✓
[ship]→ api · monitoring on
done · in production · owned by you
0
scoped proposal, back to you
weeks
to a working prototype, not quarters
0
of systems shipped with evals
yours
you own the code, models & data
specialty/
Where we go deepest: document workflows.
Across the AI stack, document processing is where relatent has the most reps — turning messy PDFs, scans and emails into reliable, automated pipelines that read, decide, and write into your systems of record.
document pipelinescroll to run →
invoices
Invoice & PO processing
Match line items to orders, flag exceptions, and post straight to your ERP.
contracts
Contract intelligence
Extract clauses, obligations and dates; surface risk before anything is signed.
claims
Claims intake
Read mixed claim packets, structure them, and route by type and severity.
kyc
KYC & onboarding
Verify identity documents and forms, cross-check, and assemble a clean file.
trade
Trade & shipping docs
Reconcile bills of lading, customs forms and invoices across a shipment.
inbox
Email-to-workflow
Turn unstructured inbound mail and attachments into structured actions.
services/
The whole AI stack — strategy, build, and the part where it actually ships.
01
AI strategy & consulting
consultingWhere AI actually pays off for you — roadmaps, feasibility, architecture and honest build-vs-buy.
02
LLM applications
llmAssistants, copilots and RAG grounded in your data — accurate, evaluated and safe to ship.
03
Agentic systems
agentsMulti-step agents that plan, call your tools and complete real tasks — with guardrails and humans in the loop.
04
Fine-tuning & adaptation
finetuneAdapt open and frontier models to your domain, tone and tasks — smaller, cheaper, better at your job.
05
Classical ML & data science
classicForecasting, classification, scoring and optimisation — where a well-built model beats an LLM.
06
Production & MLOps
productionEvaluation, monitoring, guardrails and infra so it runs reliably — and your team can operate it.
work/
Put to work across the whole field of AI.
automation
Agentic automation
Agents that take real, multi-step work off your team's plate — with approval gates and a full audit trail.
data
Data & pipelines
Clean, label and wire the data your models and agents actually need — from source to warehouse to index.
forecasting
Prediction & forecasting
Demand, risk and churn models where classical ML wins on cost and accuracy — measured, not guessed.
search
Search & RAG
Semantic search and retrieval over your knowledge — hybrid, fast, and grounded with sources.
models
Custom & fine-tuned models
Models tuned to your domain, tone and tasks — smaller, cheaper and better at your job than a raw frontier call.
process/
From a brief to a system your team owns.
01
Scope
goal, data & constraints
02
Prototype
working prototype in weeks
03
Harden
evals · guardrails · monitoring
04
Handover
you own and extend it
01 ▸ Scope
We find where AI actually pays off.
The goal, the data and the constraints — and we define exactly what "done" and "good enough" mean before any code.
02 ▸ Prototype
A working prototype on your real data.
Within weeks you see value and feasibility on your own data — not a slide deck. Classical model, LLM or agent, whatever fits.
03 ▸ Harden
We make it a system you can trust.
Evaluations, guardrails, human-in-the-loop and monitoring turn the prototype into production-grade infrastructure.
04 ▸ Handover
We hand the keys back.
We deploy into your stack, document everything, and train your team to operate and extend it — no lock-in.
stack/
Works with your stack — models, data and tools.
openaianthropichugging_facepytorchvllmlangchainpgvectorsnowflakepostgress3salesforceslackrest / webhooksaws / azure / gcp
request_quote/
Let's scope your first project.
Tell us what you're trying to do with AI. You'll get a scoped approach and a quote back within two business days.
info@relatentai.com · Barcelona, Spain