Skip to content
View victormends's full-sized avatar

Organizations

@emap2021

Block or report victormends

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
victormends/README.md

João Victor Mendes

"I treat production behavior as the real specification: isolate variables, find root causes, fix structurally so the failure class cannot recur."

Typing SVG

LinkedIn   Résumé

Location   Current employer   English

Impact Highlights


PostgreSQL & Database Engineering

  • Eliminated 100% of WAL data losses across 500+ environments by querying pg_replication_slots, identifying an orphaned slot holding WAL segments for a decommissioned subscriber, and enforcing max_slot_wal_keep_size as a fleet-wide guardrail.
  • Cut restore time 60% (17 min to 6:58) on a 15GB+ database via pg_restore -j 4 parallel processing. Traced 1GB/month disk bloat through pg_class and pg_toast to encrypted paths serialized as thousands of characters per audit log entry.
  • Automated 30-database cluster recovery with a PowerShell script that parallelizes state checks, clears stale postmaster.pid files, and handles WAL recovery on boot. Recovery cut from hours to under 5 minutes, daily support calls to near-zero.
  • Restored database access fleet-wide after a pg_hba.conf SCRAM-SHA-256 vs MD5 mismatch, applying scoped MD5 by IP range while preserving SCRAM everywhere else.


Automation & Data Engineering

  • Engineered a 70M-record ETL pipeline in a single day using PL/pgSQL + PowerShell against a 15GB+ government dataset. Unlogged tables and dynamic SQL cut processing time 60%, surfacing the top 0.001% of leads that drove Q3 commercial strategy.
  • Compressed fiscal invoice diagnosis from 30+ min to under 5 min using LLM-assisted XML parsing of raw SEFAZ rejection messages across NCM, CFOP, ICMS, and CBNEF formats, eliminating an entire class of developer escalations.
  • Zero data-loss legacy migration: reverse-engineered a Firebird database with no documentation in an afternoon, replaying 20 years of transaction history (10,000+ items, 500k BRL) from first principles to reconstruct current inventory state.

Early signal: OBMEP Silver Medal (2016), Bronze (2014), 2x Honorable Mentions; top 0.01% among 18M+ participants nationally.



name:              João Victor Mendes
role:              L2/L3 Technical Support Engineer
company:           Pedroso Automação  # B2B fiscal compliance ERP SaaS
location:          União da Vitória, PR, Brazil

specialty:         [PostgreSQL, Incident Response, ETL, Windows Server Automation]
clients:           500+ corporate  # 1500+ end-users
proof_point:       OBMEP Silver Medal  # top 0.01% among 18M+ participants

featured_now:
  - "receita-cnpj-etl-pipeline  # Receita CNPJ ETL + PostgreSQL client classifier"
  - "incident-response-runbook  # PostgreSQL incident response + NVC + post-mortems"
  - "pg-incident-recovery      # Windows-first PostgreSQL multi-cluster recovery tool"

open_to:           [Support Engineering, Database Engineering, SRE]
contact:           linkedin.com/in/mendes-victor

Portfolio Map

Area Public artifact What it demonstrates
Data engineering receita-cnpj-etl-pipeline Windows-first ETL for Receita Federal CNPJ open data, PostgreSQL staging, large-file classification, sanitized release packaging
Incident response incident-response-runbook PostgreSQL severity classification, triage, communication, escalation, and post-mortem structure
Recovery automation pg-incident-recovery Safe Windows service recovery boundaries for PostgreSQL crash-recovery scenarios
Restore tooling pg-restore-tool Operator-controlled local restore flow with explicit safe/fast/unsafe profiles
Deployment automation windows-postgres-deployment-installer Human-reviewed PostgreSQL deployment automation for Windows workstations and small servers
Fiscal support utilities nfe-xml-organizer, tls-fix-erp-updater Practical PowerShell utilities around NF-e file handling and Windows TLS registry failure modes
Data visualization one-piece-ratings-timeline Static-site data product, source policy, scheduled refresh, and GitHub Pages publishing

Featured Repos

incident-response-runbook

A production-grade PostgreSQL incident response framework that combines:

  • WAL disk exhaustion recovery (pg_wal zero bytes free)
  • SCRAM/MD5 and pg_hba.conf authentication failures
  • Connection pool exhaustion and lock contention diagnosis
  • Worked post-mortems for both infra and application-layer incidents
  • NVC-based client communication and war-room protocols

This is the clearest public artifact of how I approach production systems: diagnose from first principles, document the failure mode precisely, and encode the fix so the class of incident does not recur.

pg-incident-recovery

A Windows-first PostgreSQL recovery orchestrator for multi-cluster hosts. It automates the safe parts of post-reboot incident handling:

  • discovery of stopped PostgreSQL services
  • data-directory resolution from service definitions
  • stale postmaster.pid cleanup
  • pg_controldata-based missing WAL triage
  • parallel startup with bounded retries
  • crash-recovery monitoring and clean handoff back to Windows SCM
  • persistent queue files so interrupted runs can resume cleanly

This is the executable complement to the runbook repo: the runbook explains how I reason about incidents; this tool shows how I automate one of the ugliest recovery paths without crossing into unsafe automation.

Deployment and Restore Utilities

windows-postgres-deployment-installer and pg-restore-tool cover the less glamorous support paths around PostgreSQL: workstation/small-server deployment, local rebuilds, restore profiles, and operator-reviewed automation. They are deliberately scoped tools rather than platform claims: the goal is to remove repetitive failure-prone steps while keeping destructive choices explicit.


Technical Stack

PostgreSQL   PowerShell   Python   AWS   Git   Windows Server   Bash

WAL Management PL/pgSQL EXPLAIN ANALYZE Replication Slots Bloat Analysis autovacuum

ITIL RCA P1/P2 Incident Response ETL Pipeline Design NF-e / CT-e / SEFAZ pg_hba / SCRAM / RBAC


Education


FGV — Data Science & Economics (2020-2022) Full-ride OBMEP scholarship



IFPR — Systems Analysis & Development (2023-2024) Grade A in Mathematics for Computing


Public artifacts are sanitized: no client data, private hostnames, credentials, or internal support tickets.


footer

Pinned Loading

  1. receita-cnpj-etl-pipeline receita-cnpj-etl-pipeline Public

    Windows-first PowerShell ETL pipeline for Receita Federal CNPJ open data using PostgreSQL.

    PowerShell

  2. brazil-ncm-table-tools brazil-ncm-table-tools Public

    Windows-first .NET and PowerShell tools to download and convert Brazil's official NCM table for ERP/search workflows.

    C#

  3. pg-incident-recovery pg-incident-recovery Public

    Windows-first PostgreSQL recovery orchestrator for multi-cluster hosts, with bounded retries, crash-recovery monitoring, WAL triage, and persistent state.

    PowerShell

  4. incident-response-runbook incident-response-runbook Public

    Production-grade incident response framework for PostgreSQL environments that treats communication failure as a first-class technical problem.

    1

  5. one-piece-ratings-timeline one-piece-ratings-timeline Public

    Fan-made One Piece ratings timeline grouped by saga, with IMDb-sourced ratings

    PowerShell

  6. windows-postgres-deployment-installer windows-postgres-deployment-installer Public

    Windows-first PostgreSQL deployment automation for operator-reviewed workstation and small-server installs.

    PowerShell