param087

Use when preparing raw data for modeling — handling missing values, duplicates, inconsistent types, outliers, and bad categorical values. Emphasizes fitting all imputation on train-only to avoid leakage.

Use when running ML experiments that need to be compared, reproduced, or shared. Covers MLflow/Weights & Biases logging, what to track, run organization, and model registry basics.

Use when starting on a new dataset, before modeling, or when asked to "explore", "profile", "understand", or "summarize" data. Covers structured EDA, distributions, correlations, target leakage checks, and visualization.

Use when creating, encoding, scaling, or selecting features for ML models. Covers categorical encoding, numeric transforms, datetime/text/aggregation features, and leakage-safe target encoding.

Use when optimizing model hyperparameters. Covers search strategy (random vs Bayesian/Optuna), leakage-safe tuning inside CV, search-space design, early stopping, and budget management.

Use when the target is rare (fraud, churn, disease, anomalies). Covers correct metrics, resampling (SMOTE/undersampling), class weights, threshold tuning, and avoiding the accuracy trap and resampling leakage.

Use when fine-tuning a large language model. Covers choosing full vs LoRA/QLoRA, dataset formatting, the transformers/PEFT/TRL stack, key hyperparameters, and evaluating fine-tunes without overfitting.

Use when a model won't learn, loss is NaN, metrics look too good/bad, or training is unstable. Provides a systematic decision tree for diagnosing data, optimization, and generalization failures.

Use when building API endpoints or server actions — validate every input with zod, return consistent typed JSON errors, set correct status codes, and authenticate/authorize before touching data.

Use when adding sign-in/sign-up or session handling to a SaaS app — set up Auth.js (NextAuth) or a hosted provider with secure sessions, OAuth + email, and server-side session checks that protect routes and actions.

Use when controlling what a user may do — implement role-based access control with a central permission map, server-side checks on every mutation, and a typed `can()` helper instead of scattered role string comparisons.

Use when work shouldn't run in the request path — offload emails, webhooks, exports, and scheduled tasks to a queue or job runner with retries, idempotency, and dead-letter handling instead of blocking the user.

Use when querying or mutating the database from the app — wrap Drizzle in server-only, tenant-scoped functions instead of calling the ORM from components, so every read/write is typed, testable, and isolation-safe.

Use when designing the Postgres schema or setting up Drizzle ORM — model users, organizations, memberships, and tenant-scoped tables with the right keys, indexes, timestamps, and migrations from day one.

Use when shipping a SaaS app to production — set up CI checks, preview deploys, environment/secret management, database migrations on release, and the health checks and rollback path that make deploys boring.

Use when wiring up env vars, secrets, or runtime config — validate them with zod at boot, split server-only from public (NEXT_PUBLIC_) values, and fail fast with a clear error instead of a cryptic undefined at runtime.

Use when building or reviewing interactive UI that must meet WCAG 2.2 AA — focus management, ARIA roles and states, keyboard navigation, labels, and visible focus. Covers the accessibility mistakes agents make most.

Use when building authentication UI — login, signup, forgot/reset password, email verification, and OAuth buttons. Covers layout, validation, security-conscious UX (no user enumeration), and loading/error states.

Use when building pricing tables, plan cards, paywalls, and upgrade prompts. Covers monthly/annual toggles, highlighting the recommended plan, usage limits, feature gating UI, and clear upgrade CTAs.

Use when structuring React components in a SaaS app — composing shadcn/ui, cva variants, compound components, controlled vs uncontrolled state, and knowing when to abstract versus inline.

Use when building a dashboard or overview page — stat/KPI cards, chart placement, responsive grids, and a scannable hierarchy. Covers metric cards with deltas, where charts go, and loading/empty states for data.

Use when building data tables — sorting, filtering, pagination, row selection and bulk actions with TanStack Table + shadcn/ui, plus the loading, empty, and error states tables almost always forget.

Use when setting up a design system — color, typography, and spacing scales as CSS variables, the shadcn/ui theme, and dark mode. Covers semantic tokens, Tailwind wiring, and consistent theming across components.

Use when handling the non-happy UI states — skeleton loaders, empty states with a clear CTA, error states with retry, and optimistic updates. Covers the states agents forget that make an app feel broken or polished.