Create and publish
Normalized product information and generated Mercado Livre API payloads through a controlled workflow.
Sanitized professional case study
Catalog automation with controlled AI.
A production web dashboard that automated Mercado Livre listing creation, publishing, editing, and auditing while keeping AI-generated output behind validation and human review.
Context
From repetitive catalog work to one operational dashboard.
Large-catalog operations involved repetitive publication work, inconsistent content, slow updates, and limited visibility into catalog-wide issues. The solution centralized those workflows while preserving explicit review points.
Edit published listings
Retrieved and updated existing listings from a centralized interface instead of repeated manual edits.
Audit the catalog
Consolidated listing information so incomplete records and inconsistencies could be reviewed at scale.
Sanitized architecture
Integrations coordinated around a reviewable workflow.
Operator
Web dashboard
Backend orchestration
Validation and review
Mercado Livre API
Supporting integrations included OAuth2, cloud storage, Google Drive, Seller7, and AI-assisted image and listing generation. Deployment topology and proprietary data structures are intentionally omitted.
AI-assisted engineering
Acceleration with evidence, not blind acceptance.
AI supported both product workflows and software development. Generated output was treated as a proposal that had to satisfy explicit constraints and verification criteria.
- Structured prompts and explicit context
- Reusable skills for repeatable tasks
- Task-specific validation harnesses
- Guardrails and business-rule checks
- Automated tests and security checks
- Human review before publication
AI can accelerate implementation, but evidence determines whether the result is accepted.
Results
More than 500 listings processed through a maintainable foundation.
Centralized operation
Creation, publication, editing, and auditing were brought into one dashboard.
Reviewable automation
Repetitive steps became controlled workflows with explicit validation points.
Practical AI governance
AI generation remained behind harnesses, guardrails, tests, and human review.
Versão resumida em português
O MLBot foi um painel web de produção que automatizava a criação e publicação de anúncios no Mercado Livre, além de oferecer edição de anúncios publicados e conferência geral do catálogo.
Projetei e implementei a solução de forma independente, integrando OAuth2, armazenamento em nuvem, Google Drive, Seller7 e recursos de IA. O sistema foi utilizado no processamento de mais de 500 anúncios.
A IA era aplicada de forma controlada, com prompts e contexto estruturados, skills reutilizáveis, harnesses de validação, guardrails, testes automatizados, checagens de segurança e revisão humana.