No-code model training
Upload a spreadsheet with text and category labels, or a ZIP of pictures organized into category folders. StayCharted AI Model Trainer learns the categories from examples without requiring prompts or data-science work.
StayCharted AI Model Trainer
Freemium
No-code text and image classification trained on your labeled business examples.
O que é StayCharted AI Model Trainer?
StayCharted AI Model Trainer is a no-code platform for training classification models on labeled business examples. It accepts text in spreadsheets and images in category folders, then assigns those categories to new records such as support tickets, product descriptions, expense rows, or photos. The workflow is intended for recurring classification decisions where a team already has examples of the desired outcomes.
Before training, StayCharted AI Model Trainer checks examples for issues including duplicate rows, overlapping categories, insufficient examples, personal information, conflicting image labels, near-duplicate pictures, and unreadable image links. It evaluates models on held-out examples, reports category mix-ups, and returns a confidence score with each answer. Results can be reviewed by people, filled into CSV or Excel files, downloaded, or returned through the API on eligible plans.
O que StayCharted AI Model Trainer pode fazer?
Multiple model approaches
Choose word-based Classifier Models for clearly worded categories, pre-trained AI Classifier Models for meaning and varied wording, or Dedicated AI Models on Business for complex text and subtle distinctions. Vision add-ons support image classification.
Pre-training data checks
Review duplicate rows, categories with similar meanings, categories with too few examples, personal information, conflicting image labels, near-duplicate pictures, and unreadable links before training.
Accuracy testing and confidence review
Test models on examples they did not see during training, inspect category mix-ups, and use confidence scores to route uncertain answers to a Review Queue for human confirmation or correction.
File and API workflows
Fill CSV or Excel files in the app, download completed files, or call published models through the API. The API returns a category label and confidence for text or picture-link inputs on eligible plans.
Sensitive-data controls and team workspaces
For text models, detected emails, phone numbers, payment-card numbers, IBANs, IP addresses, passwords, API keys, and credentials in links can be masked, excluded, or retained. Workspaces include Builder, Reviewer, and Viewer roles, unlimited members, and an audit log.
Casos de uso
“Route support tickets”
Train on previously categorized customer requests, then assign incoming tickets to support queues. Teams can use confidence thresholds to send uncertain routing decisions to a reviewer.
“Standardize product catalogs”
Map supplier descriptions or inconsistent product names to approved internal catalog categories, even when new descriptions use different wording with an AI Classifier Model.
“Apply internal expense codes”
Use historical invoice lines or expense rows labeled with company codes to classify new records and produce a filled file for downstream processing.
“Sort product, damage, or inspection photos”
Provide labeled picture folders or image links to train an AI Image Classifier Model, then categorize new photos while flagging uncertain results for review.
“Connect classification to business systems”
Publish a trained model and send individual records from code, spreadsheets, or automation tools through the API. The service returns a label and confidence for each prediction.
Perguntas frequentes
What data can StayCharted AI Model Trainer learn from?
It learns from examples your team has already labeled. Text models use spreadsheet rows containing text and categories; image models use pictures arranged in category folders or image links with labels.
How does StayCharted AI Model Trainer evaluate a model?
It tests the model on held-out examples that were not used during training, shows how often it was correct, identifies categories it mixes up, and indicates how many results may need human review.
How are uncertain results handled?
Each answer includes a confidence score. Results below a confidence cutoff can be sent to a Review Queue, where a team member confirms or corrects them. Corrections can inform a later model version.
Can models be used outside the StayCharted AI Model Trainer app?
Yes, API access is included with Essentials and Business. Published models can be called using GET /api/v1/models, POST /api/v1/models/{id}/classify for one record, or POST /api/v1/models/{id}/predict for one record or up to 100 records.
What are the API throughput limits?
The API is intended for a steady stream of records. The rate limit is 30 records per minute on Essentials and 60 on Business across the whole workspace; requests that exceed the limit are refused with a 429 response and do not consume the allowance.
Informações rápidas
- Product type
- No-code text and image classification platform
- Input formats
- Labeled spreadsheet text, CSV or Excel files, picture folders in ZIP files, and labeled image links
- Model outputs
- Category label and confidence score; filled files can be downloaded
- API base URL
- https://amt.staycharted.com
- Workspace plans
- Free, Essentials, Business, and Enterprise
- Team access
- Unlimited members with Builder, Reviewer, and Viewer roles
Análise de tráfego de StayCharted AI Model Trainer
Os dados de tráfego são apenas para referência.
- Classificação do domínio
- 31