Projects Portfolio

A selection of projects demonstrating cloud architecture, security, automation, and AI infrastructure.

This portfolio is split into two sections: AI & Personal Projects showcase passion and self-driven exploration in emerging technologies, while Work Projects represent real-world solutions delivered for production environments under budget and compliance constraints.


Self-Hosted AI Infrastructure & LLM Cluster

Dual-node local LLM cluster with Hermes Agent orchestration and OpenWebUI for ChatGPT-style conversations.

Overview

Built to explore whether self-hosted infrastructure could deliver cloud-comparable inference performance while maintaining full data sovereignty and reducing cloud costs. This cluster runs production-grade LLM workloads entirely on-premises.

System Architecture

AI Cluster Architecture Primary Node (Supermicro X9DAi) 🖥️ Dual Intel Xeon E5-2697v2 💾 192GB RAM 🎮 3× NVIDIA RTX 3060 12GB 🤖 Ollama + Hermes Agent Secondary Node (Gigabyte B550) 🖥️ AMD Ryzen 5 5600X 💾 32GB RAM 🎮 5× NVIDIA RTX 3060 12GB 🤖 vLLM + Hermes Agent 🌐 Network Layer (172.18.255.0/24) Docker Compose · Docker Networks · Docker Registry Containerized Services · Port Management 🤖 Hermes Agent Orchestration Layer Multi-Agent Task Execution · Tool Use · Agent-to-Agent Coordination

Figure 1: High-level architecture of the AI cluster infrastructure

Request Processing Flow

AI Request Processing Flow OpenWebUI User Interface ChatGPT-style Hermes Agent Orchestration Multi-Agent Ollama Local LLM Inference vLLM High Throughput Serving → 📥 Ingest Logs, Data 🔍 Analyze Pattern 🛡️ Detect Threats 📊 Respond Actions, Alerts

Figure 2: End-to-end AI request processing workflow

Server Configurations

Primary Node - Supermicro X9DAi
🖥️ Hardware Specifications
  • Dual Intel Xeon E5-2697v2 CPUs
  • 192GB RAM
  • 3× NVIDIA RTX 3060 12GB GPUs
🎯 Use Cases
  • Development & testing
  • Agent experimentation
  • Mid-sized model inference
⚙️ Configuration

View supermicro-compose.yml

📦 Docker Services
  • Ollama - LLM inference engine
  • Hermes Agent - Orchestration
  • OpenWebUI - Web interface
Secondary Node - Gigabyte B550
🖥️ Hardware Specifications
  • AMD Ryzen 5 5600X CPU
  • 32GB RAM
  • 5× NVIDIA RTX 3060 12GB GPUs
🎯 Use Cases
  • Large model inference
  • Parallel workload execution
  • Production workloads
⚙️ Configuration

View gigabyte-compose.yml

📦 Docker Services
  • vLLM - High-throughput serving
  • Hermes Agent - Orchestration
  • OpenWebUI - Web interface

Technology Stack

🧠 AI/ML Stack
  • Ollama - Local LLM inference
  • Hermes Agent - Multi-agent orchestration
  • OpenWebUI - Conversational interface
  • vLLM - High-throughput serving
  • Llama.cpp - Model inference engine
⚙️ Infrastructure Stack
  • Docker - Containerization
  • Docker Compose - Service orchestration
  • Docker Networks - Service discovery
  • NVIDIA CUDA - GPU acceleration
  • Python - Automation & scripting

Security Use Cases

Log Analysis

Automated analysis of security logs using local LLMs for pattern detection and anomaly identification.

Threat Summarization

Automated threat intelligence summarization and prioritization for faster incident response.

Detection Content

AI-assisted drafting of detection rules, signatures, and security content for SIEM systems.

SOC Advisor

AI-powered advisor workflows that mirror SOC analyst patterns for faster threat investigation.

Data Sovereignty

Full data control with on-premises processing, eliminating cloud data exposure and reducing costs.


Work Projects

Production solutions delivered for enterprise clients, government agencies, and healthcare organizations. Each project represents real-world constraints including compliance requirements, budget limitations, and stakeholder expectations.

Federal SIEM & Log Architecture

Enterprise-scale log ingestion and detection pipeline for Google Security Operations (Chronicle) across 9 AWS accounts.

Client: Federal Government Agency

Challenge: The client needed a unified security monitoring architecture across multiple AWS accounts, with consistent log ingestion, compliance-aligned retention, and integration into Google Chronicle for threat detection and hunting.

Key Features:

  • Standardized log ingestion across 9 AWS accounts into a centralized Log Account
  • Chronicle forwarders for Palo Alto firewalls and Tenable VMS
  • Log retention as code — Terraform modules enforce compliance-aligned retention by default
  • FedRAMP and NIST 800-171 aligned log architecture
  • CI/CD pipeline optimization — internal S3 bucket for runners and TFSEC scanner

Stack: Terraform · Google Security Operations (Chronicle) · Palo Alto · Tenable · GitLab CI/CD · AWS CloudWatch · S3 · IAM

GitLab Repository: gitlab.com/gcp_projects3/siem-log-architecture


Healthcare Cloud Security Platform

End-to-end encrypted, HIPAA-aligned cloud infrastructure for PHI/PII data with automated cost optimization.

Client: ConcertAI (Healthcare Data Platform)

Challenge: Deliver secure, compliant cloud infrastructure for healthcare clients handling sensitive patient data, with full accountability for security architecture, automation, and cost optimization.

Key Features:

  • End-to-end encryption for PHI/PII data environments
  • CloudWatch + Lambda cost optimization — automated shutdown of non-production resources on weekends/holidays
  • SOAR-style automation for incident response and least-privilege enforcement
  • Kubernetes (EKS) cluster management and security hardening

Stack: AWS EKS · GitHub · Jenkins · Terraform · Ansible · CloudWatch · Lambda · Python · IAM · S3 ACLs

GitLab Repository: gitlab.com/gcp_projects3/healthcare-platform


Cloud Network Automation Suite

Python automation framework for Cisco router configuration, SSL certificate lifecycle, and Direct Connect failover.

Client: Ellucian Inc. (Higher Education SaaS)

Challenge: Eliminate a 2-week deployment bottleneck for hybrid cloud router configuration and automate SSL certificate lifecycle management across the enterprise platform.

Key Features:

  • Python automation generating Cisco 2900-series router configs — deployment time reduced from 2 weeks to 24 hours
  • Automated SSL certificate lifecycle — DigiCert API → S3 → AWS ACM/ELB deployment
  • Direct Connect on Catalyst switches with multi-region failover (US-EAST-1 / US-WEST-2)
  • Full-mesh Transit Gateway architecture for inter-region encrypted connectivity

Stack: Python · Cisco IOS · Catalyst Switches · AWS Transit Gateway · DigiCert API · ACM · ELB · Okta

GitLab Repository: gitlab.com/gcp_projects3/network-automation


Muller's Garage Website

Bilingual (English/Portuguese) static website hosted on Google Cloud Storage with GitLab CI/CD pipeline.

Client: Muller's Garage

Challenge: Design and deploy a responsive, bilingual website for a small business, hosted on Google Cloud Storage with automated deployment from a dev-to-main branch workflow.

Key Features:

  • Brazilian flag color palette (green #009C3B, yellow #FFDF00, blue #002776)
  • Bilingual support: English and Portuguese
  • Google Cloud Storage hosting with custom domain
  • GitLab CI/CD pipeline with dev→main branch workflow
  • Fully responsive design

Stack: HTML · CSS · Google Cloud Storage · GitLab CI/CD · gsutil

GitLab Repository: gitlab.com/gcp_projects3/lfallc.us