DriftPainel
Monitor drift between training data and current data to decide when to observe, keep or retrain models.
Includes statistical profiling, global score, feature ranking and no_action, monitor and retrain states.
Espírito Santo, Brazil · open to remote opportunities
LUCAS GARCIA / AI · DATA · SOFTWARE
Lucas Garcia: data, AI and software.
33 audited repositories and 8 real case studies with verifiable code, decisions, metrics and limitations.
DATA INTELLIGENCE PRODUCT
SYSTEMS MAP / 01
From data to product
Select an atlas region
A.01P.12
INTELLIGENCEModels, RAG, baselines and evaluation.
SELECTED SYSTEMS / 01—08
Projects connecting the whole stack
Each system record starts with the problem and ends in public links, stack, evidence and limitations verified in the repository.
AI Data Quality Checker
Analyze CSV files repeatably, reveal quality issues and generate practical cleaning recommendations.
Detects nulls, duplicates, cardinality issues, inconsistent types, empty and constant columns, problematic names and outliers.
Portal Paroquial Integrado
Organize parish administration, pastoral workflows, public communication and governance with secure module-based access.
Implemented vertical flow: create person, view person, edit person and audit changes.
Pass-To-Pass
Answer operational incidents from local runbooks with diagnosis, checklist and traceable sources.
Delivers diagnosis, resolution steps, copyable checklist, sources, confidence, feedback and operational metrics.
VagasInLine
Collect, normalize and analyze international remote jobs from public sources and structured files.
Includes Bronze/Silver/Gold, Data Quality, Schema Drift, Data Lineage, CSV/Parquet/Markdown/HTML exports and a safe scraping policy.
Previsor de Renda
Estimate monthly income from socioeconomic and occupational features with methodological rigor and interpretability.
Documented evaluation on 25,801 observations: CatBoost reached 0.5913 log-scale R² and the hybrid model reached 0.4906 R² on the original scale.
SmartClass IaC
Turn a manual full-stack deployment into declarative, repeatable infrastructure on AWS.
The repository contains 96 files, three Terraform modules and Docker images for frontend and backend, plus the full-stack application.
Daily Learning Mission
Turn continuous-learning goals into grounded weekly missions without repeating topics, duplicating deliveries or relying on opaque model decisions.
Four workflows, 149 nodes, five tables, 25 functional tests and 1,021 invariants; the full production cycle remains explicitly pending.
08 / 33 selected projects / audited repositoriesView full inventory ↗
SYSTEMS MAP / 01
From data to product
Select an atlas region
A.01P.12
INTELLIGENCEModels, RAG, baselines and evaluation.
FIELD NOTES / EXPERIENCE
Public projects, documentation and technical delivery.
The focus is making evaluation easier for recruiters: open code, READMEs, pipelines, tests and objective case studies.
Experience ↗- 01Sep 2025 - Aug 2026
Scientific Initiation Fellow in Artificial Intelligence
Ifes · Campus Colatina · PRPPG/PICTI
- 02Volunteer extension
Software Quality and Testing at LEDS
LEDS · Ifes
- 032026
Income Predictor (Undergraduate Capstone)
Ifes · Campus Colatina
RESEARCH GARDEN / NOTES
Learning documented, not hidden.
RAG and AI evaluation for intellectual-property assets
A record of scientific-initiation work combining information retrieval, LLMs and experimental evaluation in the intellectual-property domain.
- RAG
- LLM
- PICTI
- Information Retrieval
Drift as an operational routine, not a generic alarm
How global score, feature ranking and history help decide between monitoring and retraining.
- MLOps
- Dados
- Drift
Data quality as a product experience
Scores, recommendations and reports reduce the distance between technical diagnosis and action.
- Data Quality
- FastAPI
- Vue
Responsible ETL for job intelligence
Controlled collection, manual review and lineage protect pipeline quality.
- ETL
- DuckDB
- Lineage
OPERATING PRINCIPLES
How I work
Evidence before opinion
Decisions get metrics, logs, tests and limits.
Architecture serves the problem
Complexity enters when it proves value.
End-to-end reliability
Data, APIs, interfaces and documentation share quality.