degz@arbiz
System boot // loading
Open to AI engineering roles LLM applications Evaluation harnesses Automation infrastructure Remote · APAC · EU · US overlap

Hello, I'm Degz Arbiz

AI Systems Developer for LLM apps and automation

I build production AI systems, LLM evaluation harnesses, prompt architecture, and business automation.

AI Systems Developer and automation architect. LLM-powered applications, the evaluation harnesses that keep them honest, and the workflow infrastructure that connects them to real business operations. 8+ years across digital systems, SaaS and automation.

80 / 285 Sales Coach AI commits AI layer owned on a six-person team Remote APAC · EU · US overlap
AGENT-READY · curl my résumé ↗

01 / Hire me

Hire me for

Scoped engagements, senior execution, no hand-holding. Everything below is something I've shipped in production.

Calabarzon, PH · remote worldwide APAC · EU · US overlap Full-time · contract · fixed-scope

02 / Proof in numbers

Shipped, measured.

80 commits authored on Sales Coach AI — of 285 total, the largest single share
116 file changes in packages/ai, the layer I own
8+ years across digital systems, SaaS and automation
7 industries served — realty, finance, auto, government, education, hospitality, coaching
2 commercial SaaS tiers built and operated
6–7 fig client launches supported with system-driven growth models

Free live AI development training, run weekly · co-founder of Scale With System

03 / Work

Selected builds.

Sales Coach AI — AI pipeline layer

Apr – Aug 2026

Real-time sales coaching over live calls: a TypeScript monorepo with a Fastify API, Next.js client using live tab capture, and a BullMQ worker running Claude pipelines against call audio, backed by Postgres, Redis and R2. Built by a team of six; I own the AI layer — the coaching pipelines and prompt architecture in packages/ai — and wrote the cue-evaluation harness that gates output against a scripted quality bar instead of manual review. 80 of 285 commits, the largest single share.

RoleAI layer owner on a six-person team
Scopepackages/ai and cue-evaluation harness
Evidence80 of 285 commits · 116 AI-layer file changes
Claude APITypeScriptPrompt architectureLLM evalFastifyNext.jsBullMQDrizzle
Product views · scroll to explore
Sales Coach AI live coaching view: cue cards surface a competitor battlecard and a price-objection reframe while the prospect transcript streams alongside.
Live coaching — cues fire against the streaming transcript
Sales Coach AI in critical-only mode, suppressing lower-priority cues during a call.
Critical-only mode — cue volume tuned per rep

BranDoor — AI brand identity studio

2026 · in development

An independent brand identity studio that turns a short business questionnaire into a strategy brief, distinct visual directions and editable vector marks. I built the Next.js product, its staged AI workflow, a library of thirteen sample identities and photographic case studies that show each brand in use. The interactive studio is currently a local prototype; these screenshots show the product and sample work without suggesting a public demo is live.

Next.jsTypeScriptPrismaAI workflowBrand systems
Product views · scroll to explore
BranDoor homepage with its introduction and invitation to explore brand identities.
BranDoor — brand identity studio homepage
Dayfold sample brand case study showing its wordmark on coffee packaging and a takeaway bag.
Dayfold — one of thirteen sample brand case studies

Scale With System CRM

Ongoing

Built and operate a white-labelled sales and marketing platform for the Philippine market — CRM and pipeline, appointment scheduling, funnel builder, course portal, unified inbox, and local payment integration. Two commercial tiers with live paying customers.

GoHighLevelSaaS operationsPaymentsProduct

04 / History

Eight years of systems.

Oct 2025 —
Present

AI Systems Developer

Freelance / Self-employed
  • Design and build AI-powered applications on Claude and other LLM APIs, translating business requirements into prompt architecture, backend workflows, and scalable AI logic.
  • Engineer role-based, constraint-driven instruction frameworks to improve model reliability and consistency.
  • Integrate AI systems with APIs, CRMs, and automation platforms; test and optimise outputs for production accuracy.
  • Architect reusable AI components and modular workflow systems, documented for maintainability.
Ongoing

Co-founder & Systems Lead

Scale With System — KMH Web And Tech Solutions
  • Co-run a consultancy helping agencies and digital businesses scale through automation, CRM strategy, and AI-enhanced funnel infrastructure.
  • Supported multiple 6–7 figure client launches with system-driven growth models.
  • Built and operate a commercial CRM/marketing SaaS across two pricing tiers.
Sep 2024 —
Sep 2025

Automation & AI Technical Specialist

Freelance / Self-employed
  • Built lead-nurturing and sales follow-up automation using GoHighLevel and LLM tooling.
  • Developed end-to-end training and recruitment systems; managed membership and onboarding workflows.
Sep 2025 —
Feb 2026

Healthcare Advocate, Tier 2

Conduent
  • Tier 2 support for complex healthcare benefits, billing, and prescription drug inquiries under regulatory compliance requirements.
2020 —
2024

Online Business Manager · Ecommerce Copywriter

Freelance / Self-employed
  • CRM administration, funnel and website builds, and email marketing for coaching, consulting, and debt-relief clients.
2019 —
2020

Consultant · Customer Coach

Cloudbeds — remote
  • Onboarded and supported hotel properties through SaaS implementation; authored documentation for the product knowledge base.
2011 —
2018

Earlier

Open iT · TELUS International · Teleperformance · Radyo Natin

Inside sales, technical support, and customer service across enterprise software and BPO — where the systems-and-process instinct came from. Full detail in the résumé.

05 / Stack

Tools I ship with, daily.

ai
Claude API · LLM orchestration · prompt architecture · model behaviour control · evaluation harnesses · AI coding agents
automation
GoHighLevel · HubSpot · n8n · Make · Zapier
build
Python · TypeScript · JavaScript · Node · Fastify · Next.js · REST APIs · webhooks
data
Postgres · Drizzle · Redis · BullMQ
infra
Docker · Railway · Cloudflare R2 · Playwright

06 / Teaching

I teach this, live, every week.

A free live training taking complete beginners from an idea to a working, deployed AI application — turning a concept into requirements an AI can act on, building with AI coding agents, then running, testing and iterating. No programming background required.

Live weekly Non-programmers welcome Idea → deployed app AI coding agents

07 / Clients

What they said.

They went well beyond the web and funnel work — taking on responsibilities closer to a project manager.
Jonathan E. — CEO, Step Up To Care
I can now take vacations without any worries.
Marielle T. — CEO, Good Diva Marketing
The consultations were valuable enough that I offered to pay before the project even began.
Yvonne R. — CEO, Investahomes Realty

08 / FAQ

What teams usually ask.

What exactly do you own on a project?

The AI layer end to end: prompt architecture, the pipelines that call the model, the integration points into your APIs and CRM, and the evaluation that proves output quality before it ships. On Sales Coach AI that's packages/ai plus the cue-evaluation harness.

I work well alongside an existing frontend and platform team — I don't need to own the whole stack to be useful.

How do you keep LLM output reliable?

Structured input/output contracts, role-based and constraint-driven instruction frameworks, and a scripted evaluation gate rather than eyeballing results. If output quality drops below the bar, the build fails instead of quietly shipping.

Do you work with non-technical teams?

Routinely. Most of my client work has been translating what a business owner wants into something a system can actually do. I also run a free weekly live training taking complete beginners from an idea to a deployed AI app, so explaining this clearly is a practised skill, not an afterthought.

Your background is automation and BPO, not a CS degree. Why hire you for AI?

Because the hard part of production AI is rarely the model. It's knowing which business process is worth automating, what the failure modes cost, and how to integrate with the messy tools a company already runs. Eight years of doing exactly that — CRM systems, funnels, support operations, SaaS onboarding — is why my AI work ships instead of demoing.

The engineering is real and inspectable: TypeScript monorepos, job queues, evaluation harnesses, containerised deploys.

What are you available for?

Contract, fixed-scope, fractional, or full-time. Based in the Philippines and used to working across APAC, EU and US hours. Email is fastest — I reply within a day.

09 / Agents

Read me from your agent.

This site publishes machine-readable surfaces, so automated screeners and AI agents can parse my background without scraping the page.

bash
$ curl degz12348-tech.github.io/resume.txt

DIEGO "DEGZ" ARBIZ JR.
AI Systems Developer / Automation Architect
Calabarzon, Philippines · remote · APAC / EU / US overlap

10 / Contact

Need someone who ships AI from spec to production?

Open to contract work, fractional AI engineering, and full-time roles. Email is fastest — I reply within a day.