I'm a Linux Systems Engineer passionate about building and automating infrastructure at scale — from bare-metal to cloud, from Proxmox clusters to ML pipelines.
With years of hands-on experience in infrastructure and DevOps, I design, deploy, and maintain production environments spanning SDN controllers, ZFS storage, container orchestration, and monitoring at scale — while currently expanding into Machine Learning Engineering through a postgraduate program at FIAP (PosTech).
My day-to-day involves turning complex infrastructure problems into automated, observable, self-healing systems.
Core interests:
- Infrastructure as Code & GitOps
- Observability & Predictive Monitoring
- ML Engineering applied to Operations (AIOps)
- Network Automation & SDN
Group project from the FIAP PosTech ML Engineering postgraduate program. Applied Machine Learning to real-world problems as part of the Tech Challenge curriculum — covering the full lifecycle from data exploration to model deployment.
Infrastructure and DevOps challenge project showcasing skills in CI/CD, containerization, and cloud-native deployments.
The all-in-one Magic: The Gathering companion app — built by a player, for players.
- Problem: MTG players juggle 3–5 different tools just to manage their collection, build decks, track prices, and find local events.
- Solution: VaultSpell brings everything into one place — collection manager, AI-powered deck analysis, real-time price tracking (USD & BRL), wishlist, trade/sell marketplace, and a local events & stores directory.
- Stack:
React·Node.js·Scryfall API·AI/LLM·PostgreSQL - Status: Beta — actively developed, new features shipped weekly.
- Available in 🇧🇷 Portuguese · 🇺🇸 English · 🇪🇸 Spanish
Projects built in production environments with proprietary code — happy to discuss the approaches and architectures behind them.
- Problem: Reactive storage management — teams only noticed ZFS pool exhaustion when alerts fired at critical thresholds.
- Solution: End-to-end ML pipeline that pulls historical Zabbix telemetry, runs linear regression (30/14/7-day windows with R² validation), and predicts days-until-full. Integrated with n8n for automated reporting.
- Stack:
Python,Zabbix API,n8n,Linear Regression,Parquet,Ollama/LLaMA.
- Problem: No centralized visibility into circuit distribution across SDN gateways; overloaded gateways went undetected.
- Solution: Automated workflow polling the WNC SDN controller API, tracking per-gateway circuit counts with stateful thresholds, and alerting via Rocket.Chat when gateways exceed capacity.
- Stack:
n8n,REST API,Rocket.Chat,JavaScript.
- Problem: Manual node provisioning with inconsistent configurations across sites (network bonds, Corosync, LDAP, ZFS pools).
- Solution: Ansible-driven automation for full node lifecycle — from network interface configuration (LACP bonds, VLAN-aware bridges, Corosync links) to FreeIPA/LDAP integration and ZFS pool creation. Idempotent playbooks covering 10+ sites.
- Stack:
Ansible,Proxmox API,Bash,FreeIPA,ZFS.
- Problem: Basic ZFS monitoring only caught failures after the fact; no early warning for degradation patterns.
- Solution: Extended telemetry wrapper scripts feeding a six-axis severity model (capacity, pool integrity, fragmentation, scrub health, snapshot health, ARC performance) into Zabbix, with LLM-generated analysis summaries.
- Stack:
Bash,Zabbix,Ollama/LLaMA,n8n.
- Problem: Open tickets aging silently in the backlog with no visibility for the team.
- Solution: Automated weekly JQL queries detecting tickets open >7 days, with formatted card-style notifications pushed to Rocket.Chat including direct links and metadata.
- Stack:
n8n,Jira API,Rocket.Chat Webhooks.
💼 LinkedIn: linkedin.com/in/janotijr
🐙 GitHub: github.com/Janoti
🃏 VaultSpell: vaultspell.com