Skip to content

The Team

Aninda Sundar Howlader edited this page Jul 20, 2026 · 1 revision

The team of specialists

Behind GRU953-Studio there is a team of 38 specialist roles — a project lead, builders, testers, a security checker, writers, designers and more. That sounds like a lot to manage. The good news is that you never manage any of it.

This page is here purely if you're curious. You never need to remember a single name on it.

You only ever talk to one person

You talk to one friendly coordinator: the Project Lead. It listens to your idea, asks you a few clear questions, and then quietly brings in whichever specialists your particular project needs. Everyone else works behind the scenes. When their work is done, the Project Lead gathers it up and gives you one plain-English reply — never a pile of technical chatter.

Each project also wakes only the roles it actually needs. A tiny website wakes just a handful. A bigger app — one with logins, saved data, and its own pictures and words — wakes more. You don't decide who joins; the Project Lead does, and it always leans towards the smallest team that can do the job well.

If you haven't set it up yet, the Installation page walks you through it step by step.

How to read the tables below

Each table has three columns:

  • Role — the specialist's name.
  • Runs on — which member of the Claude family of AI models does the work. There are three: Haiku is the quickest and cheapest, Sonnet is the balanced middle, and Opus is the most capable (and the most expensive). More on why this matters in Why roles run on different models below.
  • What it does — in plain words.

Core team (14)

Most projects use these.

Role Runs on What it does
Project Lead Opus The one coordinator you actually talk to. Runs the whole project, brings in the right specialists, and gives you a single plain-English reply.
Interviewer Sonnet Prepares the multiple-choice pop-up questions that get your idea clear before any work starts.
Architect Opus Suggests a few ways to build your idea, then breaks the chosen one into small, checkable steps in the right order.
Scope Guardian Sonnet Stops extra features quietly creeping in, and keeps a note of anything you decided to leave out.
Builder Sonnet Writes the actual code, one small piece at a time.
Reviewer Opus Reads the finished code independently to catch mistakes before they reach you.
Tester Sonnet Proves each part really works — and never claims success without running the checks.
Security & Compliance Auditor Sonnet Scans for leaked passwords, known weak spots and licence problems, and checks any personal data is handled properly — before anything ships.
Brand Guardian Haiku Keeps the look, colours and wording consistent with your brand.
AI Developer Sonnet Adds AI features safely, only when you genuinely need them, with sensible safeguards built in.
Fixer Sonnet When something breaks, quietly tries a small fix first, and only troubles you if that doesn't work.
Cost Monitor Haiku Keeps an eye on running cost and favours the cheaper option wherever it can.
Publisher Sonnet Chooses the version number, writes the plain-English release notes, and ships it privately to your own GitHub.
Memory Keeper Haiku Remembers everything between sessions, so you can stop and pick up exactly where you left off.

Content team (5)

These join in once you approve the mock-up (the clickable preview of your app). They fill it with its real words and media. All five run on Sonnet.

Role Runs on What it does
Content Director Sonnet Plans all of the app's real content from your approved mock-up, and looks after the one-time media opt-in.
Text Content Specialist Sonnet Writes the genuine words inside your app, in Bangla and English — real writing, not machine translation.
Image Content Specialist Sonnet Creates the app's images, icons and illustrations, each with alt-text (a short written description for screen readers) — using your own Google account, always asking first.
Audio Content Specialist Sonnet Creates any audio or narration, with written transcripts — using your own Google account, always asking first.
Video Content Specialist Sonnet Creates any video clips, with captions — using your own Google account, always asking first.

A specialist for every platform and language (10)

Whichever kind of app you want — for Android, iPhone, Mac, Windows, Linux or the web — there's a specialist who knows that platform's tools inside out. The Architect picks the right one for you. All 10 run on Sonnet.

Role Runs on What it does
Flutter/Dart developer Sonnet The cross-platform default — one build that runs on phones, tablets, and more.
Kotlin developer Sonnet Native apps for Android.
Swift developer Sonnet Native apps for iPhone, iPad and Mac.
Java developer Sonnet Java apps and services.
C# developer Sonnet Apps for Windows and the .NET platform.
Python developer Sonnet Python apps, scripts and tools.
Rust developer Sonnet Fast, memory-safe Rust programs.
Go developer Sonnet Services, command-line tools and Linux programs, written in Go.
C++ developer Sonnet High-performance C++ programs.
TypeScript developer Sonnet Websites and web apps.

Brought in only when your project needs them (9)

These wake up only when your particular project calls for them.

Role Runs on What it does
Maintenance Agent Sonnet Later fixes and new features on a project you've already published — picking it up safely, rather than starting over.
DevOps Engineer Sonnet Puts your app online, and can add standard self-recovery so a live app quietly restarts itself if it ever crashes.
Responsible-AI Reviewer Opus An independent fairness and honesty check on any AI feature that affects a real decision about a person.
UX Designer Sonnet Makes the app pleasant and clear to use.
Accessibility Specialist Sonnet Makes sure people with disabilities can use it too.
Technical Writer Sonnet Writes a plain guide for your finished app.
Data Engineer Sonnet Designs safe, reliable storage for your app's data.
Localisation Specialist Sonnet Adds more than one language, for example English and Bangla.
Researcher Sonnet Checks current facts, and can suggest an existing Claude Code tool when one already solves the job — always asking before installing anything.

Why roles run on different models

You'll have noticed that different roles run on different models. That's on purpose, and it keeps your costs down.

The rule is simple: each job uses the cheapest model that can do it reliably. Only the hardest reasoning is given to the most capable — and most expensive — model.

  • Haiku handles the tidy, mechanical work where there's little to weigh up: keeping the brand consistent, watching costs, and looking after the project's memory.
  • Sonnet is the balanced everyday choice, and does the bulk of the work — building, testing, writing content, and each of the platform specialists.
  • Opus is held back for the genuinely hard thinking: running the project (Project Lead), designing it (Architect), reviewing the code for correctness (Reviewer), and judging whether an AI feature is fair and honest (Responsible-AI Reviewer).

A second opinion from Ollama (free to run)

Several roles can also ask for an independent second opinion from Ollama — a free-to-use AI model (no cost) that runs privately on your own computer, without the cloud. The roles that can do this are the Reviewer, the Security & Compliance Auditor, the Architect, the Builder, the DevOps Engineer, and the Publisher.

It's always your choice, and the team asks before installing anything. Claude stays the default. See Connecting Ollama for how it works and how to turn it on.

The totals

To confirm the numbers: there are 38 roles in all — 3 run on Haiku, 31 on Sonnet, and 4 on Opus. You never operate any of them by hand. The Project Lead coordinates the lot, and only ever wakes the ones your project needs.

Related pages

You can also read the project's full README or visit the GRU953-Studio website.

Clone this wiki locally