Speech
ASR, speaker labeling, multilingual audio, transcription, and structured speech datasets.
Datasets
Coverage across the data types modern AI products depend on.
Dragon AI supports both shelf-ready and custom-built datasets so customers can start quickly or scope a program around a specific model, domain, or deployment need.
Text
Instruction data, QA corpora, classification sets, knowledge assets, and language data for advanced AI learning.
Image
Classification, detection, segmentation, captioning, and visual taxonomy annotation.
Video
Clip tagging, event detection, temporal segmentation, scene understanding, and multimodal alignment.
Audiovisual
Synchronized audio-video datasets for multimodal training, annotation, and evaluation workflows.
OCR and Documents
Document parsing, key information extraction, and layout-aware annotation.
Multimodal
Image-text, video-text, and speech-text collections for aligned systems.
Fine-tuning
Supervised fine-tuning data, prompt-response pairs, and instruction tuning sets.
Preference and RLHF
Ranking data and pairwise judgments for model alignment.
Evaluation
Benchmark creation, adversarial testing, and release-stage regression suites.
Delivery Options
Choose shelf-ready data for speed or custom programs for specific model goals.
Off-the-Shelf Datasets
Best when teams need to move quickly, test feasibility, or accelerate prototyping with structured data that is already available.
Custom Dataset Programs
Best when data requirements involve unique domains, languages, policies, taxonomies, or evaluation criteria.