Research to Product

We turn applied research
into AI products

We create products and services from research in statistics and AI. Each delivery states the problem it addresses, how it was tested, and which limitations remain.

See the products Talk to us
3 AI Products
PhD + MSc in Statistics
14 Years of Experience
In production

See the method in motion

In 32 seconds, see how Lumina works: create a version, check it against stated criteria, decide, and improve.

AI-generated presenter (Théo), a declared synthetic character.

About Us

From research
to market

Lumina is an AI product lab. We design, operate, and document products built from statistical and AI research. We also use them in our own work to compare what we promise with how they perform and with public sources.

Founded by a PhD and a Master's in Statistics, the company has roots in applied scientific research, statistical consulting for healthcare institutions, and corporate data science experience. This academic foundation evolved into an innovation lab focused on generative AI.

We use a "Research to Product" workflow: define a problem, build a testable product, operate it, and revise it from evidence.

Scientific Foundation

Each product states which public sources and statistical assumptions guide its decisions

Evidence-Led Delivery

Small releases, explicit criteria, and review before a change becomes publicly available

Monitored Use

We state how the product works, what we monitor, and which limitations remain

Clients & reference projects
A2GO Claude Code workshop
Hospital Albert Einstein Hospital Albert Einstein Statistical consulting
Colégio Brasileiro de Radiologia (CBR) Colégio Brasileiro de Radiologia Statistical consulting
100+ Graduate Students Personalized statistical consulting across multiple universities
The DoesIt family

The DoesIt family

Every product in the DoesIt family follows the same process: define a question, test under controlled conditions, and record what the results support. Each product applies that process to a different problem.

The family keeps growing — more DoesIt on the way

Courses & Skills

Learn to build evidence

Five bilingual courses: the eight-track Full Harness and four standalone courses. AI Coding OS is listed separately as a companion skills pack.

Course · 8 tracks

The Judge-Driven Harness

Understand the failure. Choose the check. Decide from evidence.

One course organized into eight tracks: CORE teaches the method and seven tracks apply it. The core path starts without code; runnable labs are available for readers who also want to implement the method.

FalhasEvidênciaDecisãoPráticaPT/EN
8tracks
58modules
57labs
See the course and its tracks
Skill pack

AI Coding OS

Shared rules and checks for coding agents.

A companion pack separate from courses and tracks: 18 installable SKILL.md resources for Claude Code, Cursor and Codex.

Claude CodeCursorCodexSKILL.mdEN/PT
3IDEs
1contract
18resources
See the pack
Applied course

Applied Statistics for Health Research

Read tables. Understand the method. Interpret the result.

Learn the basics for following analyses in health articles, understanding why a method was chosen, and performing simple comparisons. Every topic includes an example and guided practice.

ArticlesTablesInterpretationGuided practiceEN/PT
13modules
14labs
2runtimes
Explore the course
Expertise

AI Solutions

Areas of expertise we combine according to the problem, the data, and the delivery format

01Statistical consultingResearch, healthcare and business decisions 02AI agent developmentTools, RAG, integrations and control 03Predictive modelingForecasting, risk and machine learning 04AI evaluation and governanceEvaluation, security, and monitoring
See all services and delivery formats →
Explore technical capabilities

Generative AI

AI Chat Intelligent chatbots with memory and context
Agentic Flows Autonomous agents with multi-step orchestration
RAG Retrieval-Augmented Generation on your data
Web Search Integrated web search with AI synthesis

Evaluation & Governance

LLM Evals Systematic evaluation of quality and accuracy
Guardrails Security, filters and compliance for LLMs
MLOps & Observability Deployment, monitoring and quality gates for models in production
Training & Fine-tuning Custom models for your domain

Developer Tools

AI Code Generation Generative AI-assisted code generation
AI Code Review Automated PR and quality reviews
AI-Powered CI/CD Smart pipelines with testing and deployment

Data & Modeling

Time Series Forecasting, seasonality and temporal embeddings
Custom Models Custom statistical and ML models
Data Science & Statistics Advanced analysis, survival, tests and inference
Statistical Genomics Statistical modeling of genomic data and livestock genetics
Healthcare Analytics AI for clinical data, oncology and public health

Education & Training

Statistics for Research Applied statistics for scientific research — developed by PhD and MSc researchers in Statistics
Custom Training Programs Tailored programs for teams and institutions — from fundamentals to advanced methods
Methodology

Research to Product

Our process turns a research question into a product that can be tested and used

01

Research

We start with a concrete problem, check what public sources support, and define a hypothesis that can be tested.

Example Documented question + public sources
02

Prototyping

We build a small version, define criteria, run tests, and document the limits we observed.

Example Problem → prototype → reproducible test
03

Product

A version that meets the criteria receives monitoring, support, and integrations suited to its intended use.

Example Reviewed evidence → release decision
Contact

Let's talk

Tell us what you need to analyse, build, or teach. We can respond with a product, consulting engagement, or training suited to the problem and your team's context.

Company Lumina Serviços de Consultoria LTDA
CNPJ 57.150.909/0001-43
E-mail info@lumytics.com
Location Santa Bárbara d'Oeste — SP, Brasil

By submitting, you authorize us to use the data to answer your request, as described in our Privacy Policy.