ISA

Engineering

Engineering

My machines, the tools I code with, how I work with AI, and the experiments I keep alive on the side. A living page, not a resume.

Current stack

Setup

Categories you can explore. Each one opens a focused view of the tools and devices in it.

Machine

The hardware I actually work on every day: a Windows desktop for heavy work and local models, a MacBook Air for portable coding, plus the peripherals and devices around them.

9 itemsOpen ->

Editor

Where I write code. AI-native editors and IDEs I switch between depending on the project.

3 itemsOpen ->

AI Coding Workflow

Terminal agents, AI pair-programming tools, and the subscriptions that power them. This is how most features actually get built.

5 itemsOpen ->

Local Models

Local LLMs I run on the RTX 4060 Ti for automation, research, and article-generation experiments.

2 itemsOpen ->

Featured

Highlighted gear and tools

What I rely on most day-to-day.

Claude Code

Active

Terminal agent for shipping features step by step

Roadmaps, file-level changes, and larger feature work

Cursor

Active

AI-native editor for in-context coding

Editing with AI assistance inside the codebase

Main PC

Active

Main Windows workstation and local-LLM rig

Coding, Windows development, local model testing, experiments

qwen3:32b

Testing

Larger local model for research and writing tests

Local automation, research, and article-generation experiments

Codex

Active

Terminal and app coding agent I use daily

Plan-driven feature work in the terminal and app

MacBook Air M1

Active

Portable machine for daily coding and writing

Portable development, daily coding, writing, admin work

Workflow

How my coding flow has evolved

From writing every line by hand to working alongside agents.

  1. Step 01

    Manual Coding

    Writing code by hand to actually learn the language and the problem. Java and C/CMake school work started here; it is still my baseline for understanding.

  2. Step 02

    AI-Assisted Coding

    Explaining the feature or bug to an AI tool first, asking for a roadmap and file-level changes, then implementing and testing step by step.

  3. Step 03

    Terminal Agents

    Using Codex and Claude Code in the terminal to ship larger features in steps, while still reading and understanding the diffs before they land.

  4. Step 04

    Local LLM Experiments

    Testing Qwen and Gemma locally for automation, research, and source-based article generation, prototyping agent-like workflows off the cloud.

  5. Step 05

    Review and Cleanup

    Once the feature works, I go back and improve structure, move hardcoded content into admin-editable data, and tidy the parts I rushed.

Log

Engineering log

Short technical updates. Lab notes, not full posts.

  • Engineering section with admin management

    Categories, gear cards, workflow, log, experiments, and principles, all editable from /admin/engineering instead of hardcoded.

    May 2026Shipped

  • Cinema and Radio sections

    Personal scoring, notes, and metadata lookups for films and music on the blog.

    May 2026Shipped

  • Association Management System

    Association/community tool: admin features, room/space management, PDFs, notifications, backups, and tests. Still in active development.

    Ongoing

  • Academic article automation

    Source-based local-LLM writing pipeline that drafts academic-style articles from provided sources. Prompts and structure still being tuned.

    Testing

  • Local LLM workflow on the home GPU

    Running qwen3:32b and gemma 3 12b locally to prototype automation and agent ideas without cloud cost.

    Testing

  • Personal blog structure cleanup

    Consistent cards, shared primitives, and moving more content behind the admin panel.

    Ongoing

  • Local LLM Discord agent

    A small assistant in Discord backed by a local model. Scoped but not started.

    Planned

  • Automation dashboard

    One place for personal scripts, schedulers, and small services. Idea stage.

    Planned

Experiments

What I'm building or testing

A grid of in-flight ideas, each with a quick status.

Association Management System

Association/community management app: admin features, room and space management, PDF generation, notifications, backups, and testing. Tags: TypeScript, Next.js, admin, PDFs.

Active

Academic Article Automation

A source-based writing system where local LLMs draft academic-style articles from provided references. Tags: local LLM, automation, research, writing.

Testing

Local LLM Research Assistant

A focused tool for reading, summarizing, and cross-linking notes, running fully on local models. Tags: local LLM, research, Qwen, Gemma.

Testing

Local LLM Discord Agent

A small assistant that lives in Discord and answers against a local model. Tags: local LLM, Discord, agent, automation.

Planned

Personal Blog CMS

The admin surface that powers this site, so content is editable without touching code. Tags: Next.js, admin, content, local-first.

Active

Multi-LLM Comparison Workspace

A workspace to send the same prompt to multiple models and compare answers side by side. Tags: LLM, comparison, tooling, TypeScript.

Testing

Automation Dashboard

A single board for personal scripts, schedulers, and small services. Tags: automation, dashboard, tooling.

Planned

Principles

How I like to build

A short list I come back to when a decision feels heavy.

  • 01Ship fast, then clean. Get it working, then improve the structure.
  • 02AI first, but understand the code. I don't merge what I can't explain.
  • 03Prefer simple architecture over clever architecture.
  • 04Build tools I actually use, not demos.
  • 05Make content admin-editable instead of hardcoded when possible.
  • 06Local-first when it makes sense; reach for the cloud when it doesn't.
  • 07Design matters, but function comes first.
  • 08Keep personal projects alive instead of starting new ones.