Di Lucaro
Booking Q3 — Stockholm / RemoteTwo-person consultancy

Production AI systems designed for reliability.

Di Lucaro is a Stockholm consultancy pairing applied AI engineering with senior systems architecture. We design and ship AI systems for teams that can't afford for them to fail.

[ 01 / What we help with ]

Where we're useful.

We take on engagements where AI reliability matters — regulated industries, internal tools that have to stay correct, and products where “mostly works” is a business problem.

Enterprise RAG

01

Hybrid retrieval, reranking, and evaluation harnesses for AI search that survives real users and regulated environments.

Hybrid searchEval harnessOn-prem

Document intelligence

02

Pipelines that extract structured data from invoices, contracts, and technical documentation — with audit trails, not just outputs.

ExtractionValidationAudit

Agentic workflows

03

Autonomous agents with deterministic guardrails, confidence routing, and human-in-the-loop review where it matters.

OrchestrationGuardrailsHITL

Systems architecture

04

Senior systems and infrastructure design for the platform the AI runs on — distributed systems, integrations, and long-running services.

ArchitectureIntegrationScaling

Reliability audits

05

Independent review of an existing AI system: retrieval quality, prompts, evals, observability, and the failure modes you haven't seen yet.

AuditEval reviewRoadmap

AI visibility & content

06

Retrieval-aware content architectures, thought-leadership programs, and LinkedIn strategy — how a brand surfaces inside AI answers and the writing that gets it there.

GEOArticlesAnalytics
[ 04 / Philosophy ]

How we think about reliable AI.

Reliability patterns from firmware

01

State machines, circuit breakers, watchdog timers, checkpointing. AI systems need the same discipline as embedded code.

Evaluation-first design

02

If you can't measure it, you can't ship it. We start with the eval that defines success — not the prompt that looks good in a demo.

Retrieval quality > model size

03

Hybrid search, re-ranking, and good chunking beat bigger models. Most production wins come from the data pipeline.

Compound reliability

04

A 95% step in a 5-step pipeline is 77% end-to-end. We design with the chain in mind, not the demo.

[ 05 / Who works here ]

Two senior engineers. Stockholm.

Applied AI engineering paired with senior systems architecture. Manning Publications technical reviewers, financial markets infrastructure experience, and a firmware-trained reliability mindset.

Meet the team