melissa-pereira-deel

Data Builder — Data & AI agent for Claude Code

Data & AI community

Builds and audits datasets for tiny models — synthetic generation, teacher labelling, held-out splits, and label-quality checks.

How to install Data Builder

Installs to ~/.claude/agents/melissa-pereira-deel-tiny-model-lab-data-builder.md

Terminal
mkdir -p ~/.claude/agents && curl -fsSL https://raw.githubusercontent.com/melissa-pereira-deel/tiny-model-lab/HEAD/.claude/agents/data-builder.md -o ~/.claude/agents/melissa-pereira-deel-tiny-model-lab-data-builder.md

Restart Claude Code, or start a new session, for it to be picked up.

What Data Builder does

name: data-builder description: Builds and audits datasets for tiny models — synthetic generation, teacher labelling, held-out splits, and label-quality checks. Use after triage returns a tiny-model verdict and before any training run. tools: Read, Write, Edit, Bash, Grep, Glob, WebSearch model: opus

You build the dataset. For tiny models this matters more than architecture, and it is where most projects quietly fail.

Source hierarchy, best first

  1. Deterministic teacher. An ex

Alternatives in Data & AI

  • Silas — Database Architect 296 ★
  • Triage TS — LLM instruction for the TypeScript orchestrator triage role 93 ★
  • PyTorch — PyTorch ML model implementation, training loops, and model export 65 ★

Full documentation available on GitHub

View Source Repository

Related Agents

Task Triage — Data & AI agent for Claude Code

Task Triage

Decides whether a task should be a tiny model at all. Use FIRST, before any data collection or training, whene

Data & AI community
Budgeter — Data & AI agent for Claude Code

Budgeter

Resource gatekeeper. Other agents must call the budgeter before launching long-running training, expensive LLM

Data & AI community
Data Quality Auditor — Data & AI agent for Claude Code

Data Quality Auditor

Audits datasets and pipelines for the silent failures — grain violations, join fan-out, freshness gaps, reconc

Data & AI community
Judge Writer — Data & AI agent for Claude Code

Judge Writer

Writes a binary LLM-as-judge prompt for exactly one failure mode, from a definition and a handful of human-lab

Data & AI community
Ava Issue Triage — Data & AI agent for Claude Code

Ava Issue Triage

Triages open GitHub issues — ranks them by real user impact rather than by label, verifies claims against the

Data & AI community
Agent — Data & AI agent for Claude Code

Agent

AI video production orchestrator. Routes requests to specialized ComfyUI skills for character generation, vide

Data & AI community

Related Skills

Claim Audit — Security skill for Claude Code

Claim Audit

A Claude skill that audits AI-written text claim by claim: quotes each checkable claim, tags it with one of te

Security community
Jev Curate — AI skill for Claude Code

Jev Curate

High-throughput synthetic & pretraining dataset sifter powered by TypeSafe AI Jev (api.typesafe.ai). Stream, f

AI community
Cv Scientist Skills — Security skill for Claude Code

Cv Scientist Skills

Ten Claude Code skills for image and video ML — data audit, splits, preprocessing, augmentation, model choice,

Security community