Skip to content

Data Science · Applied AI · Automation

Turning legacy systems into business decisions

I'm Marcelo Santos. I build data systems, automation and artificial intelligence for real-world operations.

From ERP integrations to autonomous agents, semantic search and large-scale monitoring, I design every solution to fit each company's business rules.

Not prototypes. Production systems — running processes, connecting information and turning data into action.

View projects Let's talk
Sheet 01 — Ecosystem in production click a system
Read as
ERP UAU The heart of the companySQL Server + API Platform AppTech All management in one placeFlask · React · ~20 ETLs IA · RAG ai-memory The memory AI looks upHybrid search · MCP 15 agents Robot Fleet Routines that run themselvesScheduled jobs · daemons Orchestrator Pandora Asks before it signsAuditable agents · LLM Messaging AutoMgsWeb The ERP notifies who needs itIdempotent queue · 3 DBs Agent Collections Robot Reconciles and settles alone6-stage pipeline Platform Court Monitor Only what is each client’s1.1M publications/day TOTVS RM Spreadsheets Emails · 58k AI assistants Gazettes · 91 courts Tax XMLs Court databases Managers · WhatsApp WhatsApp Clients · email ERP UAU core Platform AppTech IA · RAG ai-memory 15 agents Robot Fleet Orchestrator Pandora Messaging AutoMgsWeb Agent Collections Platform Court Monitor TOTVS RM Spreadsheets Emails AI assistants Gazettes (91) Tax XMLs Managers WhatsApp Clients
Systems7 in production
Coreconstruction ERP
Scalenot to scale · schematic
Rev.2026-09

01 Areas of expertise

Areas of expertise

Every skill points to the production systems that prove it.

Area AppTechPandoraCourt Monitorai-memoryRobot FleetCollectionsAutoMgsWeb
01

Data Science & ETL

Resilient pipelines that extract, transform and unify data from legacy ERPs, SQL Server databases, TOTVS RM and spreadsheets — with incremental loads, watermarks and replica resilience.

Python · PostgreSQL · SQL Server · ETL incremental · pandas · Airflow-style jobs

~2,7M processes in the historical pipeline — AppTech

Proven in
AppTech Pandora Court Monitor ai-memory Robot Fleet Collections AutoMgsWeb: not applicable
02

Applied AI & RAG

Semantic search and RAG systems in production: embeddings, full-text, rank fusion and an MCP server acting as a "second brain" over 58,000 emails and thousands of documents.

RAG · Embeddings · Busca híbrida · MCP · LLMs · Bancos vetoriais

58k+ emails indexed — ai-memory

Proven in
AppTech: not applicable Pandora Court Monitor: not applicable ai-memory Robot Fleet: not applicable Collections: not applicable AutoMgsWeb: not applicable
03

Automation & Autonomous Agents

Robots that run the ERP on their own: they approve payment processes, import invoices, monitor publications from 91 courts, reconcile payments and settle instalments — and notify on WhatsApp whoever needs to know.

Agentes autônomos · WhatsApp API · Integração de APIs · Jobs agendados

15 autonomous robots in production — Robot Fleet

Proven in
AppTech Pandora Court Monitor ai-memory: not applicable Robot Fleet Collections AutoMgsWeb
04

Full-Stack & DevOps

Complete web apps in Flask + React, packaged in Docker and served in production with Docker Swarm, Traefik, Nginx and rolling deploys with healthchecks.

Flask · React · Docker Swarm · Traefik · Nginx · REST APIs

13 permission-controlled modules — AppTech

Proven in
AppTech Pandora Court Monitor ai-memory: not applicable Robot Fleet: not applicable Collections: not applicable AutoMgsWeb

mark = the system delivers this skill · click to open the case

02 Projects & Case Studies

Projects & Case Studies

Production systems solving concrete business problems.

All projects

Flagship case

AppTech — Plataforma de Gestão

A satellite platform unifying sales, receivables, tax, legal and HR for a construction company.

processes in the historical pipeline
~2,7M
permission-controlled modules
13
incremental ETL pipelines
20+
Full-StackData ScienceAutomation
Read the full case
Fig. 01 — Data flow 12 nodes · 11 edges
UAU (SQL Server) TOTVS RM over VPN Spreadsheets Court databases public + one paid Incremental ETL watermark + natural-key UPSERT Enrichment rate limit + cost ceiling PostgreSQL Flask API 13 modules React SPA WhatsApp PDF / XML ERP filings queue with heartbeat
Source External Process Store Application Output

More than twenty incremental pipelines consolidate the source systems into Postgres; on top of it, thirteen modules and the outputs operations actually use — including legal filings back into the ERP. Payroll comes in over a VPN, opened only during extraction. Drawing a single Postgres is a simplification: it stands for AppTech’s own database plus the legal database it reads and enriches.

Systems index

02 Pandora — Agente de Operações do ERP An agent orchestrator that operates the ERP — and, where money is involved, asks before it signs. Applied AIAutomationData ScienceFull-Stack 8 min 2026 03 Monitor Judicial — Plataforma Multi-cliente Reads the official gazettes of 91 courts every night and delivers to each client only what is theirs — as a PDF, shielded against false positives. AutomationData ScienceFull-Stack 1,1 mi+ 2026 04 ai-memory — Segundo Cérebro (RAG) A persistent-memory MCP server with semantic search over APIs, code, emails and schemas. Applied AIData Science 58k+ 2025 05 Frota de Robôs — Operação Autônoma do ERP 15 robots in production that operate the ERP on their own — tax, receivables, legal and infrastructure. AutomationData Science 15 2024–2026