01 — output
Define what must exist
Name the working artifact first: the report that has to land, the dashboard someone opens, the workflow that has to fire. That's the target.
ai architect taking on builds
Systems that keeps working after everyone logs off.
I'm Seigfred, an AI architect. I design AI systems end to end — the data model, the agents, the pipelines, and the product around them — then ship them to production, where they turn manual, repetitive operations into systems that run on their own.
- 4 AI platforms architected
- Multi-tenant SaaS
- Voice AI agents
- End to end — design to production
workflow/client-reporting.json
executing
trigger
Webhook·n8n
process
Claude·AI Agent
store
Supabase·SQL
Done
--:--:--waiting for trigger…
1,204 runs
99.9% uptime
~3.2s avg
// selected work
Systems built, shipped, and still running.
A few of the things I've designed end-to-end — from the trigger that kicks them off to the report that lands in someone's inbox.
ai saas platform
live
Dojo Roleplay
A multi-tenant SaaS that lets martial-arts schools run realistic AI voice roleplays, auto-scores every call with OpenAI, and manages staff, scenarios, analytics and plans across schools.
View project
ai creative platform
live
Dojo Content Machine
A self-serve creative studio for martial-arts schools — upload a logo, uniform and dojo photos, pick a guided brief, and get on-brand images and ads generated in the background, credit-metered, and saved to a private gallery.
View project
ads automation
in build
AI Ads Platform
Self-serve Meta ads for martial-arts schools — generates ad creatives from the school's own references, publishes campaigns to Meta, then imports every lead and follows up with an AI conversation over SMS.
View project
agency operations
live
GrowPro OS
The agency's internal OS — ClickUp tasks land as reviewed drafts, then a resumable pipeline builds HighLevel sub-accounts step by step. Two more departments run social posting to Social Planner and AI roleplay training for client success.
View project
// the toolchain
The tools I reach for.
Chosen for reliability and how cleanly they hand off to each other — most of my systems are two or three of these wired together.
01orchestrate
n8nWorkflow orchestration
Make.comVisual automation
02ai & llm
OpenAIModels & scoring
ClaudeReasoning & agents
03data & backend
SupabasePostgres & auth
AirtableStructured ops data
Node.jsCustom logic & APIs
04interface
ReactFront-end & dashboards
05client & deploy
GoHighLevelClient CRM & comms
RenderDeployment
// how i work
I build backward from the output.
Most automation projects stall because they start from the tools. I start from the thing that has to exist, then trace the path to it. Output, then input, then process.
02 — input
Work back to what it needs
Identify the data, triggers, and knowledge the output depends on, and where each piece actually lives today — not where I wish it lived.
03 — process
Build the path between
Wire the pipeline, automate the handoffs, and make it observable — so when something breaks, it's obvious where, and the system keeps running until then.
// the human node
I'm an AI architect based in the Philippines. I design the system before anyone writes a line — where the data lives, what the AI is trusted to decide, where a person stays in the loop, and what happens when a step fails — then I build it and ship it. Most of that work happens in an agency context, where the goal is always the same: take work that's being done by hand and turn it into something that runs on its own.
That instinct started on the mat. I spent years competing in martial arts — learning to read a situation fast and trust a framework I'd drilled until it was automatic. More recently I took it to the stage, competing in impromptu speaking at the national level: building a structured argument in the minute before I walk on, no script, under pressure. Both are the same move as the systems work — a reliable framework you can lean on when it counts.
Away from all that I run and hike, and I track the numbers while I'm at it, because I'd rather measure a thing than guess at it.
RoleAI Architect
Based inPhilippines · GMT+8
FocusAI platforms, agents & pipelines
Core stackOpenAI · Supabase · Node.js · n8n
Off-stageMartial arts, speaking, hiking
Open toAI architecture & build work
// talk to my assistant
Rather just say it out loud?
Press the button and talk through what you're trying to automate. You'll be speaking with my AI assistant — not me directly — and it can answer questions about how I work and set up a call.
the human behind it
// ai assistant
ready
Tap to start talking
It listens, asks a couple of questions about your manual process, and books a call if it makes sense — no typing, no forms.
try saying
- "We copy Facebook leads into a spreadsheet by hand."
- "Our weekly client reports take a whole day."
- "Could you build something like Dojo Roleplay for us?"
what it's understood
processlistening…
how oftenlistening…
toolslistening…
fitlistening…
Want to pick this up with me?
// open a connection
Got something that should be running on its own?
Tell me what the manual version looks like today, and I'll tell you what it could look like as a system.