Available for work

I build AI automation systems that run themselves.

Full-stack automation architect. I take business problems and turn them into deployed systems - n8n workflows, Python bots, AI agents, server infrastructure. No templates. Built from scratch, built to last.

280+
Workflows Built
30
Areas Automated
40+
Containers Managed
5+
Custom AI Bots
Selected Work

Systems I've built.

Production systems running 24/7. Click "View Architecture" to see how each system works under the hood.

Meridian Bot Framework

Custom Python framework for production AI agents
Framework
Built a full async bot framework from scratch to replace slow, unreliable n8n AI Agent nodes. Meridian runs production Telegram bots with persistent memory, parallel tool execution, and multi-turn reasoning. Currently powers 5+ bots with 19+ integrated tools each.
5+
Production Bots
19
Integrated Tools
24/7
Uptime
SQLite
Session Persistence
Pythonaiogram 3.x OpenAI Agents SDKMistral Small 4 Langfuse TracingPydantic
Meridian Request Flow
Input
Telegram
User message
->
Transport
aiogram 3.x
Async handler
->
Brain
Agents SDK
Multi-turn reasoning
->
Execute
19 Tools
Parallel execution
^
Trace
Langfuse
Cost + latency
<-
Memory
SQLite
Full history
<-
Model
Mistral
Small 4
<-
Response
Telegram
Structured output

Multi-Domain Automation System

30 operational areas orchestrated through a single architecture
Live
Designed and built an automation system spanning 30 operational areas - each with its own task queues, logs, handoffs, and cross-domain communication. Handles everything from CRM pipelines and crypto trading to health tracking and content production.
280+
n8n Workflows
40+
Data Tables
30
Active Areas
0
Manual Steps
n8nBaserow REST APIsWebhooks Claude CodeCross-Domain Memory
System Architecture
Orchestrator
n8n
280+ workflows
< >
Shared State
Baserow
40+ tables
< >
Thinking
Claude Code
Session memory
serves 30 operational areas
CRM Pipelines
Twenty CRM sync
Content Production
Draft, schedule, publish
Crypto Trading
Binance automation
Health Tracking
Daily check-ins
Finance & Tax
Expense tracking
+ 25 more
All automated

Production Server Infrastructure

64GB dedicated server running 40+ containers with auto-recovery
Production
Manage a Hetzner dedicated server running the full automation stack - 40+ Docker containers orchestrated with Compose, Nginx reverse proxy with SSL, Tailscale VPN, health monitoring with auto-recovery, and multi-database management.
40+
Docker Containers
64GB
RAM Managed
99.9%
Uptime Target
Auto
Recovery
Docker/ComposeNginx Ubuntu/LinuxTailscale VPN GrafanaUptime Kuma
Infrastructure Stack
Edge
Internet / Tailscale VPN
v
Proxy
Nginx + SSL/TLS
Reverse proxy, routing
v
Docker Compose Orchestration
n8n
Workflows
Baserow
Database
Twenty CRM
x2 instances
Ollama
Local LLMs
Qdrant
Vector DB
Letta
AI Memory
PostgreSQL
+ MySQL, Redis
Listmonk
Email
+ 30 more
containers
v
Monitoring
Grafana + Uptime Kuma + Langfuse
Auto-recovery on failure

AI Memory & Autonomous Agent System

Persistent AI memory with semantic search and cross-session learning
Live
Built an AI memory layer that enables agents to remember, learn, and share context across sessions. Combines a vector database for semantic retrieval with structured memory for cross-domain findings.
600+
Memory Entries
5
Memory Blocks
RAG
Retrieval
Always-On
Orchestration
QdrantLetta EmbeddingsSemantic Search OllamaCross-Domain Context
RAG Pipeline & Memory Architecture
Input
New Data
Any domain
->
Process
Classify
Category + tags
->
Embed
Vectorise
Semantic encoding
->
Store
Qdrant
600+ entries
retrieval path
Query
Agent Request
"What do I know about X?"
->
Search
Semantic Match
Top-k results
->
Enrich
Baserow
Structured context
->
Return
Context
Informed response

Synqr - B2B Automation SaaS

AI-powered lead automation for Australian fitness businesses
In Development
Automation service for Australian fitness businesses. AI voice agents handle inbound calls, automated SMS/email nurture sequences convert leads to bookings. 10-step nurture engine with smart routing.
AI
Voice Agents
10
Nurture Steps
SMS
Auto-Response
98%
Margin
n8nTwilio VapiClaude Haiku Cal.comBaserow CRM
Lead Automation Flow
Trigger
Inbound Call
Gym's number
->
AI Agent
Vapi + Claude
Answers, qualifies
->
Capture
Baserow CRM
Lead record
v
Nurture
10-Step SMS
30-day sequence
->
Route
Smart Reply
STOP / interested / pause
->
Convert
Cal.com
Booking confirmed
Technical Skills

What I work with.

Self-taught across the full stack. Everything listed here is proven in production, not theoretical.

AI & Automation

  • n8n workflow automation (280+)
  • OpenAI Agents SDK
  • Claude / Anthropic API
  • Mistral, Groq, Ollama
  • RAG pipelines (Qdrant)
  • Prompt engineering
  • Function calling / tools

Backend & APIs

  • Python (async, aiogram, Pydantic)
  • REST API design & integration
  • Webhooks & event-driven patterns
  • PostgreSQL, MySQL, SQLite, Redis
  • Baserow (40+ production tables)
  • Data architecture & contracts

Infrastructure & DevOps

  • Docker / Docker Compose
  • Linux/Ubuntu server admin
  • Nginx reverse proxy + SSL
  • Tailscale VPN
  • Systemd services & cron
  • Hetzner bare metal

Observability

  • Langfuse (AI tracing)
  • Grafana dashboards
  • Uptime Kuma monitoring
  • GoAccess log analysis
  • Auto-recovery patterns
  • Cost tracking & optimisation
How I Work

My approach.

01

Understand the problem

I don't start with tools. I start with the business problem. What's manual? What breaks? What costs money? The system design follows from there.

02

Build for production

Everything I build runs 24/7 with auto-recovery, monitoring, and structured logging. Not prototypes - production systems that handle real data at scale.

03

Document everything

30,000+ lines of architectural documentation across my systems. Every decision, every pattern, every edge case - written down so the system outlives any single session.

04

Ship fast, iterate faster

I build and deploy within days what takes most teams months. Speed isn't about cutting corners - it's about knowing what matters, building that first, and improving from real feedback.

Let's build something.

Looking for someone who can take a business problem and turn it into a deployed, self-running system? Let's talk.

Get in touch
LinkedIn Upwork