Jayden Ontalan Capistrano Developer tooling / validation systems

Developer tooling for AI-assisted engineering workflows.

I build developer tooling for AI-assisted engineering: durable evidence records, validation scripts, guarded model calls, parser diagnostics, and self-hosted infrastructure.

View work Resume Contact
Featured
3 engineering systems
Validation
local checks + dossiers
Infra
self-hosted Linux stack
Case studies
Project Steward first

Selected work

Developer tools and evidence systems for AI-assisted engineering, validation workflows, and ambiguous technical investigation.

All projects

Evidence workflow

Turns
Evidence
Decisions

Project Steward

Developer tooling

Project Steward

Case study

A local TypeScript CLI that turns AI-assisted development work into durable records: turns, claims, evidence, reviewer results, synthesis, and decisions.

Build notes

  • Models turn records, claims, evidence, reviewer results, synthesis, and decisions as local project records.
  • Uses a guarded OpenAI reviewer path with explicit confirmation, provider gating, and audit records.
  • Keeps model output advisory; it does not automatically mutate memory, decisions, or validated project records.

System focus

  • Durable evidence records
  • Guarded model calls
  • No automatic memory mutation

Role

Developer tooling + AI workflow systems

Stage

Active build

Focus

Durable evidence records

TypeScriptNode.jsOpenAI APICLI toolingEvidence records

Updated Jul 2026

Case study

Validation tooling

Finch Standards

Case study

Local standards and validation tooling for Finch projects: required-file checks, local-only remote checks, template validation, dossier/showcase rules, and source-backed public-claim boundaries.

Build signal

Codified repo structure, Git policy, validation entrypoints, templates, issue memory, dossier coverage, and SHA-tracked bundle provenance.

ShellGitValidation scriptsProject standards

Updated Jul 2026

Case study

Systems investigation

Legacy Protocol Restoration Evidence Lab

Private lab

A private local evidence lab that structures ambiguous restoration work around architectural questions, theory deltas, run findings, and validation tiers.

Build signal

C#/.NET restoration work and Python parser/diagnostic tooling around protocol flow, runtime state gates, protocol atlas pages, and campaign verdicts.

C#/.NETPythonParser toolingRuntime diagnostics

Updated Jul 2026

Case study

Additional builds

Deployed apps, retrieval systems, forecasting tools, planning flows, and spreadsheet-native calculators that round out the project index.

SpineFrame

Production-preview, public

self-hosted reflection and memory system

Financial Forecaster

Paused, local-only

local forecasting and stress-test app

Mānoa Faculty Senate RAG

Paused, local-only

local retrieval and citation pipeline

VocabVerse

Production-preview, private beta

deployed Japanese vocabulary app

Hangout Planner

Production-preview, public

public no-login planning site

Mage's Survival Calculator

Live workbook, version-bound

formula-heavy spreadsheet calculator

How I work

The important boundary is whether a claim can be checked, traced to evidence, or held for review.

  1. 01

    Define the boundary

    Separate project records, model output, raw evidence, and open questions before automating.

  2. 02

    Build reviewable systems

    Turn workflows into durable records, validation scripts, guarded model calls, and explicit decisions.

  3. 03

    Validate before promotion

    Keep generated output advisory until local checks, evidence, or human review support the claim.

Case study focus

The three systems carrying the application story.

  • Project Steward case study: evidence-backed governance for AI-assisted development.
  • Finch Standards: local validation workflows, repo standards, dossier checks, and SHA-tracked provenance.
  • Legacy Protocol Lab: private methodology-focused systems investigation with parser/diagnostic tooling and validation tiers.

Projects, resume, or contact.

The primary recruiter path stays short: inspect the work, check the resume, then reach out directly.

Resume Contact