AI Engineer.
Agentic Systems Architect.
I build multi-agent AI systems that run autonomously on local infrastructure — zero cloud dependency, zero monthly inference costs, vendor-independent. Everything you see here was built and runs on a single M1 Mac Air
alex@saintlex:~$ system --status
[ PRODUCTION METRICS ]
> Awaiting next instruction...
Who I Am
I'm Alex Bogle — an AI Engineer and Agentic Systems Architect based in Portmore, Jamaica. I started in IT operations: troubleshooting hardware, managing 6,000+ assets, building automation workflows that cut processing time by 40%. That operational discipline is what sets me apart in AI engineering.
I taught myself AI/ML — from LSTMs to Transformers to RAG pipelines to full multi-agent orchestration. The system running this portfolio is proof: 6 autonomous agents, 26 automated workflows, 108 custom skills, 5-layer persistent memory — all at $0/month cloud cost on a single M1 Mac.
I build agentic systems with evaluation loops, RAG, and multi-layer memory — patterns I learned from Ed Donner's Agentic AI Engineer course and applied to production workflows like the content pipeline below.
I'm looking for Agentic AI Engineer, MLOps, or AI Solutions Architect roles — remote preferred. I bring real infrastructure experience, production AI systems, and the ability to ship from day one.
Professional Background
My edge is combining real infrastructure experience with AI engineering. Most AI engineers have never touched physical hardware, diagnosed a network fault, or managed a corporate asset grid. I have.
Coordinated procurement, shipping, and partner documentation for technology distribution operations. Maintained structured records, confidential partner files, and operational documents supporting business decisions for the Managing Director.
Owned remote and hands-on technical triage for provisioning teams — device imaging, OS configuration, firmware checks, network troubleshooting. Managed end-to-end lifecycle for 6,000+ assets in Asset Tiger CMDB. Reduced return-to-service lag by 35%. Built Slack workflow automations cutting processing time by 40%.
Standardized receiving, dispatch, and inventory-control SOPs — reduced fulfillment errors by 30%. Used QuickBooks Enterprise to maintain real-time inventory accuracy, reconcile discrepancies, and support audit-ready records.
Troubleshot consumer IoT hardware and app issues — Bluetooth, firmware, sensor data, biometrics, account access. Managed high-volume inbound voice and email queues with detailed CRM documentation and plain-English resolutions.
Frontline support for desktop, printer, network, email, and platform issues. Installed and maintained routers, switches, Wi-Fi, laptops, printers. Built Google Sheets automation reducing daily processing from 7 hours to 3 hours (57% improvement).
Google IT Support Professional Certificate. Systemic network architecture, terminal diagnostic automation, and access authorization — the foundation of self-healing agent infrastructure. Also: EF SET English C2 Proficient (79/100).
What I've Built
Every system below is deployed and running. Not mockups. Not tutorials. Production-grade infrastructure with real data pipelines and measurable outcomes.
Portfolio Agentic Infrastructure
6 autonomous AI agents with production delegation patterns. 52 integration tests passing. Docker Compose with FastAPI, Streamlit, Qdrant, Paperclip.
RAG Pipeline with Eval
Production RAG with LLM-as-judge evaluation. Streaming Q&A, PDF upload, MLflow experiment tracking. 6 tests, all mocked. Runs on M1 Mac at $0 cloud cost.
Hermes Agent Infrastructure
26 cron jobs, async routing, Paperclip orchestration, Obsidian knowledge management, 5-layer memory. All containerized.
Linux, Containers & Observability
Administer Linux environments for staging microservices, Docker Compose services, SSH access control, Prometheus + Grafana monitoring.
Awesome Second Brain
Contributed solution profile, setup guide, and comparison page to the leading AI second brain comparison repo. 2 PRs, 15+ files.
AgentMail for AI Agents
Full email inbox for AI agents. Send, reply, triage. CLI helper + Hermes skill. Part of the second brain stack.
AI Content Quality Pipeline
End-to-end system: draft → humanize → evaluate → retry. LLM-as-judge scoring + TextHumanize engine. Gradio dashboard for real-time comparison.
Tech Stack
Open Source Skills
Skills I've built and published for the Hermes Agent community. Installable via hermes skills install.
Open Source Contribution + Publishing
Complete workflow: discover, scope, write, submit PR, and publish across platforms (Dev.to, GitHub, LinkedIn). Includes platform picker CLI, quality test, repo-specific notes, and HermesHub submission guide.
Capabilities
Multi-Agent Orchestration
Building and deploying autonomous AI agent networks with event-driven routing, persistent memory, and async execution. LangGraph, Paperclip, Hermes.
MLOps & Infrastructure
Docker containerization, CI/CD pipelines, model serving, monitoring (Prometheus/Grafana), automated health checks and cron orchestration.
AI/ML Engineering
Transformers, RAG pipelines, fine-tuning (LoRA/QLoRA), embeddings (pgvector, nomic-embed), local inference with Ollama, evaluation frameworks.
IT Operations Foundation
4+ years hands-on: asset management (6K+), network troubleshooting, Slack/Sheets automation, Google IT Support certified, C2 English.
Writing & Research
I Processed 2.4 Billion Tokens Across 52 AI Models for $0.52
A full cost breakdown of running a production multi-agent AI system on a single M1 Mac. 26.6K requests, 52 models, $0.52 total spend.
I Built an AI Agent to Track My Job Applications
7 untracked applications found. 4 bugs diagnosed and fixed. All on a $0 cloud setup running free AI models.
Why I Built a 5-Layer AI Memory Stack on an M1 Mac
Every AI agent forgets. Here's how I built persistent, multi-layer memory that survives session restarts, API failures, and model changes — at $0/month.
I Contributed to Chroma's Open-Source Docs
Found an open issue, reviewed two existing PRs that didn't fully address it, and submitted a comprehensive rewrite. Here's what I changed and what I learned.
What Happens When You Push Free AI Models to Their Limits
Exploring the practical boundaries of free, locally-run AI models on a consumer Mac. Benchmarks, pitfalls, and lessons learned.
Hermes Agent – Zero-Cost AI Operations Blueprint
Technical deep-dive showing how a production-grade AI agent runs on a single M1 Mac with zero cloud cost.
I Built a Production RAG System on My M1 Mac for $0
Published on Dev.to. How I built a RAG pipeline with LLM-as-judge evaluation, MLflow tracking, and zero cloud cost — all on an M1 Mac in Jamaica.
I Contributed to the Awesome AI Second Brain Repo
Solution profile, setup guide, example workflow, and a new comparison page — all documenting how to build a fully local AI second brain on an M1 Mac.
How to Get Hermes Agent to Use AgentMail (And Why It's Harder Than It Should Be)
The exact issue from a Reddit user, the 3 common failure modes, step-by-step fix, and alternatives like Himalaya for Gmail users.
OpenRouter Dashboard
Live usage data from my OpenRouter management API. This is real-time data from the AI models powering my agentic infrastructure — updated on every page load.
Token Breakdown
Requests Per Day
Daily Activity
Click headers to sort| Date ↕ | Requests ↕ | Tokens ↕ | Cost ↕ |
|---|---|---|---|
| 2026-06-24 | 507 | 24.5M | $0.0000 |
| 2026-06-23 | 1.5K | 319.2M | $0.0000 |
| 2026-06-22 | 1.5K | 289.8M | $0.0000 |
| 2026-06-21 | 1.2K | 200.4M | $0.0000 |
| 2026-06-20 | 946 | 98.5M | $0.0000 |
| 2026-06-19 | 324 | 22.0M | $0.0000 |
| 2026-06-18 | 338 | 24.0M | $0.0000 |
Top Models by Requests
| Model | Requests | Tokens | Cost |
|---|---|---|---|
| openrouter/owl-alpha | 23.0K | 3.3B | $0.0000 |
| openai/gpt-oss-120b | 9.5K | 364.6M | $0.0000 |
| openai/gpt-oss-20b | 2.1K | 68.4M | $0.0001 |
| google/gemma-4-31b-it | 1.4K | 71.7M | $0.0000 |
| nvidia/nemotron-3-super-120b-a12b | 1.3K | 91.9M | $0.0019 |
| z-ai/glm-4.5-air | 78 | 3.7M | $0.0000 |
| nvidia/nemotron-3-nano-30b-a3b | 32 | 355.7K | $0.0001 |
| nvidia/nemotron-3-ultra-550b-a55b | 20 | 2.9M | $0.0000 |
| google/gemma-4-26b-a4b-it | 18 | 5.5K | $0.0000 |
| meta-llama/llama-4-scout | 15 | 691.0K | $0.0764 |
All Models (54)
54 models
Auto-refreshes every 1 hour · Last updated: 2026-06-25 02:55 UTC
Let's Talk
I'm open to Agentic AI Engineer, MLOps, and AI Solutions Architect roles — remote preferred. Based in Jamaica (ET). Production experience from day one.
Jamaica · ET · Remote-Ready
