For AI teams

Generate evaluation datasets from your documents

Upload your PDFs or spreadsheets. Get structured Q&A pairs with verified answers. Ready for your RAG pipeline.

How it works

Three simple steps to create evaluation datasets for your AI system.

1

Upload your files

PDFs, Word docs, PowerPoints, or spreadsheets (CSV, Excel)

2

Configure options

Choose question types (lookup or reasoning), language, and how many questions you need

3

Download dataset

Get a structured CSV with questions, answers, and evidence. Plug into your eval pipeline

What you get

Every dataset includes verified answers grounded in your source documents.

Example output
QuestionAnswerEvidenceType
What was the total revenue in Q4?$2.4 million"Q4 results showed total revenue of $2.4M..."Lookup
How did Q4 compare to Q3?12% increase from Q3's $2.1M"Q3 revenue was $2.1M, Q4 reached $2.4M..."Reasoning

Built for quality

Documents & Spreadsheets

Works with PDFs, DOCX, PPTX for text content. CSV and Excel for structured data.

Grounded answers

Every answer includes the exact evidence text from your source. No hallucinations.

Two question types

Lookup questions for direct facts. Reasoning questions for multi-step understanding.

Multiple languages

Generate datasets in English, Spanish, French, or Hebrew.

Domain expertise

Specialized modes for Finance, Legal, Healthcare, Tech, and HR documents.

Consistent format

Same CSV structure every time. Works with any evaluation framework.

Simple API

Upload files, create datasets, and download results. All authenticated with your API key.

  • POST /files - Upload a file
  • POST /datasets - Create a new dataset
  • GET /datasets/:id - Check status
  • GET /datasets/:id/download - Get CSV
1. Upload a file
curl -X POST api.teloset.com/v1/files \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -F "file=@document.pdf"
2. Create dataset
curl -X POST api.teloset.com/v1/datasets \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "name": "my-eval-set",
    "sourceFileIds": ["file_abc123"],
    "config": {...}
  }'

Ready to build better evals?

Start generating evaluation datasets from your documents today.

Get started for free