Skip to content
View diomari's full-sized avatar
🌍
Hello World!
🌍
Hello World!

Block or report diomari

Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
diomari/README.md

Hi, I'm Diomari πŸ‘‹

I'm a full-stack engineer building AI-enabled products, serverless systems, developer tools, and frontend experiences.

I like working at the intersection of product, architecture, and implementation: turning ambiguous ideas into scoped systems, shipping the smallest useful loop, and then hardening it with clear boundaries, tests, documentation, and operational visibility.

What I build

I focus on practical software that connects product value with strong technical foundations:

  • AI product systems β€” grounded assistants, AI agent workflows, human-in-the-loop approvals, retrieval, evaluation, cost controls, and safe automation.
  • Edge-native backend β€” Workers, Hono, D1, R2, Queues, Durable Objects, Workflows, Workers AI, AI Gateway, and serverless deployment patterns.
  • Developer tools β€” CLIs, coding-agent packages, project scanners, workflow helpers, and tools that improve engineering feedback loops.
  • Frontend products β€” React, Astro, React Native, Expo, polished UI, performance-minded animation, accessibility, and maintainable component systems.
  • Product engineering β€” scoping, architecture, requirements clarification, milestone planning, async communication, documentation, and code review.

Current themes

A lot of my recent work explores how engineers can use AI without giving up judgment, maintainability, or user trust.

I keep coming back to a few principles:

  • AI features should solve a workflow problem, not just add a chatbot.
  • Human approval is a system concern, not a button on top of an unsafe process.
  • Serverless architecture is strongest when each primitive has a clear job.
  • Good developer tools make context, constraints, and uncertainty visible.
  • Small, well-scoped loops beat impressive demos that cannot be operated.
  • Documentation and async writing are infrastructure for remote teams.

Featured projects

JigSpot

Offline-first fishing app for offshore anglers.

JigSpot is a React Native app for Filipino anglers who fish offshore, where signal is unreliable and generic maps do not show the seafloor detail needed for jigging and bottom fishing.

  • Stack: React Native, Expo, MapLibre, Cloudflare Workers, D1, TypeScript.
  • Product: Offline bathymetric maps, GPS trip tracking, spot saving, catch logging, trusted-circle sharing, and conditions history.
  • Offline-first: The phone is the source of truth during a trip; cloud sync is for sharing and backup.
  • Maps: Vector depth contours, structure edges, target-depth highlighting, and examine-area tools for depth/slope/fishability.
  • Backend: 100% serverless Cloudflare API for sync, sharing, feature gates, and account/product flows.
  • Premium model: Features such as offline downloads, premium maps, broadcast, and spot limits are gated server-side.

Key idea: For field apps, the riskiest loop is often offline behavior. The product has to keep working exactly when the network disappears.


Gorgi

A one-line-to-embed grounded AI chat widget for websites.

Gorgi is a multi-tenant AI assistant that answers from a site's own content instead of behaving like a generic chatbot. A business signs up, adds a knowledge source, drops one script tag onto its site, and visitors get grounded answers with cost controls and tenant isolation.

  • Stack: Cloudflare Workers, Hono, Workers AI, D1, React, TypeScript.
  • Product loop: Create a site, add knowledge, embed one script, ask grounded questions, control usage.
  • Frontend: React dashboard for client/site setup and source management.
  • Embed: Tiny Shadow DOM widget designed to avoid leaking styles into host pages.
  • Backend: Edge API for tenant validation, origin checks, quota checks, retrieval, answer generation, and usage tracking.
  • Data model: D1 stores metadata, site configuration, source references, and usage counters.
  • Isolation: Every request carries tenant/site context so one site never sees another site's content.

Key idea: Embeddable AI products should be boring on purpose: small payload, strict tenant boundaries, clear quotas, and predictable integration.

Top Tech Stack

  • TypeScript
  • JavaScript
  • Python
  • NumPy/Pandas
  • Node.js
  • React
  • React Native/Expo
  • HonoJS
  • NextJS
  • Astro
  • HTML5/CSS3
  • GSAP
  • Cloudflare
  • AWS
  • PostgreSQL
  • SQLite
  • AI Gateway

Pinned Loading

  1. flare-control flare-control Public

    A Cloudflare-native agent orchestration platform with a control plane, runtime plane, workflows, and admin dashboard.

    TypeScript

  2. oto-storage oto-storage Public

    Proxy-based storage wrapper that allows using localStorage and sessionStorage like a regular javascript object. Supports TTL / Expiration and Encryption

    TypeScript

  3. brief-ctx brief-ctx Public

    The missing handoff tool for AI coding agents.

    TypeScript

  4. edge-form edge-form Public

    Self-hostable Cloudflare contact form backend with spam protection, D1 submission storage, Resend/webhook delivery, an admin dashboard, and CSV export.

    TypeScript