Mark Anthony Sabandal · Agentic Engineer in the Philippines

Agentic Engineer for the layer that makes AI-assisted software dependable.

Mark Anthony Sabandal is a full-stack web developer and Agentic Engineer based in the Philippines. For more than a decade he has built and maintained software in real working environments, where a system has to do more than look finished—it needs to be dependable, understandable, and safe to operate. His focus now is the part of AI-assisted development that typing speed cannot solve: writing specifications precise enough that an agent cannot misinterpret them, designing tests and review gates that catch bad output before it ships, and supervising code-generating agents from intent to verified result.

That shows up as local-first tools, AI-assisted workflow infrastructure, media pipelines, and client-side automation. He is especially drawn to the awkward middle of a project—the handoffs, failure cases, and the small repetitive steps that make a tool frustrating in everyday use. There, he untangles the process, writes the missing structure, and leaves behind something that can be checked, picked up again, and trusted by the person using it. In professional settings, he also goes by Mark Sabandal.

Operating Thesis

The bottleneck moved. This is the layer Mark works at.

Writing code is no longer the hard part. The value now sits upstream: clear intent, trustworthy verification, and disciplined supervision of the tools that generate the code.

Specify before generating

Turn intent into tickets, plans, and requirements precise enough that an agent cannot misread them.

Test as the product

Design verification and review gates that catch bad output before anything reaches production.

Supervise the agent

Break work into agent-sized chunks, review output through packets, and correct by rewriting the prompt.

Keep the system legible

Durable artifacts, retrospectives, and runbooks preserve the knowledge a model does not have.

Operating Index

The working shape behind the portfolio.

Senior product engineering experience, local-first automation, and agentic workflow design presented as evidence rather than broad claims.

Current roleAgentic Engineer

Full-stack developer building dependable workflow systems

Long arc10+ years

Senior developer experience with SolarPlus

Build modeLocal-first

Durable state, review packets, and approval gates

Proof set6

Public-safe project groups ready for review

Selected Work

Proof-heavy project groups drawn from local repository briefs, sprint artifacts, and case-study drafts.

AI Workflow Tooling

Relayframe

Local-first control plane for turning project intent into durable plans, tickets, sprint execution, review evidence, and approval-gated release work.

  • Tickets and sprint plans act as the specification, so agents generate against explicit acceptance criteria instead of open-ended chat.
  • Review packets, test-result gates, and approval packets verify agent output before it advances.
TypeScriptNode.jsCLI

Media Automation

Twitch Content Grabber

Pipeline for turning selected Minecraft Twitch VODs into reviewed YouTube Shorts with quality gates before publishing.

  • Review and quality checks stay ahead of publication instead of after upload.
  • Scoring, captions, and metadata are grounded in transcript and correction data.
Next.jsTypeScriptPrisma

Local Desktop Automation

PhotoDrain and PhotoVault FreeUp

Local desktop assistants for Google Photos backup and cleanup with explicit confirmation before destructive actions.

  • Manual Google login avoids password storage and security bypass behavior.
  • Local validation and dedupe reporting build reviewed evidence before cleanup.
ElectronReactTypeScript

How I Build

Work is strongest when the operating path is as visible as the final interface.

01Intake
02Plan
03Build
04Verify
05Review
06Release

Case Study Track

3 priority systems have deeper drafts.

The current case-study set covers workflow infrastructure, Twitch media automation, and local desktop safety tooling.

Open case studies

Work With Mark

Start with the system, workflow, or product problem you want solved.

Send the project context, current constraints, timeline, and the best public link to review.

Contact Mark