Filling the request form

The request form captures everything our team needs to build your dataset. Required fields are marked with an asterisk.

Want a figure before you fill this in? The annotation cost estimator takes the same fields.

New Request form — schematic layout

#	Field	Details
1	Data type *	Image, Video, Audio, Text, Multimodal, or Medical data. The dropdown is searchable.
2	Data volume *	How much data you need (numerical, up to 5 digits)
3	Use case *	Computer Vision · NLP / Text Classification · NLP / NER (Entity Extraction) · Document Processing (OCR, key-value extraction) · Gen-AI dataset (prompt-output pairs, instruction tuning) · Content Moderation · Reinforcement Learning (preference data, trajectories, AI feedback) · Medical Imaging / Clinical Data · Medical Text Classification / Annotation (notes, reports, EMR)
4	Taxonomy *	Number of classes and number of objects per item
5	Annotation platform	Toggle on if you need an annotation platform set up for your project
6 · 7	Guidelines & raw data *	Upload areas for your brief and your raw dataset — covered in Guidelines & uploads

Turning on "I need an annotation platform set up for my project" notifies our team and adds a credit overhead, because it includes standing up the labeling environment for you.

Choosing your use case

The Use case dropdown tells our team what kind of model you're building, which shapes how the data is structured and labeled. Here's what each option means and the kind of training data it typically needs:

Use case	What it is	Training data typically needed
Computer Vision	Models that interpret images or video, such as object detection, image classification, segmentation, or tracking.	Images or video frames with labels: bounding boxes, polygons, segmentation masks, keypoints, or class tags per item.
NLP / Text Classification	Models that sort text into categories, such as sentiment analysis, topic tagging, intent detection, or spam filtering.	Text samples (sentences, reviews, tickets, posts) each labeled with one or more category tags from your taxonomy.
NLP / NER (Entity Extraction)	Models that find and tag specific entities inside text, such as names, dates, amounts, products, or domain-specific terms.	Text with span-level annotations: each entity marked by its start and end position and an entity type label.
Document Processing (OCR, key-value extraction)	Models that read structured or semi-structured documents, such as invoices, receipts, forms, and contracts, and pull out fields.	Document images or PDFs with transcribed text and labeled key-value pairs or regions (for example invoice number, total, date).
Gen-AI dataset (prompt-output pairs, instruction tuning)	Data to fine-tune or instruction-tune generative models, so they follow instructions and respond in the style and format you want.	Prompt-and-response pairs, instruction-input-output triples, or multi-turn conversations written or curated to your guidelines.
Content Moderation	Models that flag unsafe, harmful, or policy-violating content across text, image, audio, or video.	Examples labeled against your moderation policy: safe vs. violating, with category tags (e.g. hate, violence, spam) and severity.
Reinforcement Learning (preference data, trajectories, AI feedback)	Data to align models with human preferences (RLHF) or train agents, by comparing outputs or recording action sequences.	Ranked or pairwise preference comparisons between model outputs, reward labels, or recorded trajectories of states and actions.
Medical Imaging / Clinical Data (vision-oriented)	Vision models for healthcare, such as detecting findings in X-rays, CT, MRI, pathology slides, or other clinical images.	Medical images annotated by qualified annotators: segmentation, bounding boxes, or classifications, often with clinical context.
Medical Text Classification / Annotation (notes, reports, EMR)	Models that read clinical text, such as notes, radiology reports, and electronic medical records, to classify or extract information.	Clinical text labeled with categories or entity spans (diagnoses, medications, procedures), handled with appropriate privacy care.

Not sure which fits? Pick the closest match and describe the specifics in your guidelines. If your project spans more than one (say, document processing plus entity extraction), note that too, our team will scope it with you.