Skip to content
BigTree108

Data annotation and labelling services for AI

Data annotation and labelling services for AI: image and video annotation, text and document labelling, audio transcription, LiDAR and 3D point cloud labelling, and RLHF, preference and evaluation data for language models, with quality control on every batch, from a dedicated annotation team in your tools or a labelling project delivered to an agreed accuracy.

Annotation and labelling work we take on

  • Image annotation

    Bounding boxes, polygons, keypoints and pixel masks for detection, segmentation and pose models, drawn in CVAT, Label Studio, Encord, SuperAnnotate or Roboflow, with SAM 3 or your current model pre-labelling each image so annotators correct rather than draw from scratch.

  • Video annotation and object tracking

    Objects followed across frames with stable IDs, keyframes interpolated and corrected, and actions and events marked with start and end times, for tracking, activity recognition and video understanding models.

  • Text and document labelling

    Entities, intents, sentiment, categories and relations in text, and fields, tables and layout regions in invoices, contracts and forms, in English, Ukrainian and other languages, exported as JSONL or in the format your training code reads.

  • Audio transcription and speech labelling

    Verbatim transcripts with timestamps, speaker turns, noise and language tags, and intent and emotion labels for speech recognition, voice assistants and call analytics, with a second listener checking a sample of every batch.

  • RLHF and preference data

    Prompts written to cover your product’s real tasks, ideal answers for supervised fine-tuning, pairwise preferences and rankings for reinforcement learning from human feedback, and rubric scores, from reviewers trained on your guidelines.

  • Evaluation sets and red-teaming

    Golden question and answer sets, graded model outputs, agent traces reviewed step by step, and adversarial prompts that probe for jailbreaks, prompt injection and harmful answers, so each model or prompt change is scored before release.

  • LiDAR and sensor fusion labelling

    3D cuboids, point cloud segmentation and tracks across LiDAR sweeps, linked to the matching camera frames, for automotive, robotics and mapping datasets, labelled in Segments.ai, CVAT, Encord or your own tool.

  • Model-assisted labelling and active learning

    Pre-labels from foundation models and from your own model, active learning that sends the most uncertain items to people first, and pipelines that move data between your storage, the labelling tool and training without manual exports.

  • Quality control and gold sets

    Hidden gold items in every batch, agreement between annotators measured with Cohen’s kappa or overlap scores, a reviewer sampling each annotator’s work, and accuracy reported per batch and per label, with recurring errors turned into guideline updates.

Hire data annotators

  • Dedicated data annotators

    Data annotators who join your team full time, work in your tools and process, and report to your lead. You interview them; we carry the Ukrainian contract, payroll, invoicing and leave.

  • Data annotation projects

    A defined piece of data annotation with a scope, a fixed plan and a named lead on our side who owns the result and reports progress in your channels.

  • Ongoing data annotation

    Data annotation as a continuing service: the same people every month, a backlog you prioritise, and hours you can see in our portal and on the invoice.

The dedicated team page explains how specialists join your team, and the outsourcing page covers project delivery, take-overs and how we charge.

Who works on your data annotation

  • Data annotators

    Labelling to written guidelines in your tool

  • Annotation team leads

    Guidelines, throughput and feedback to annotators

  • Quality reviewers

    Gold sets, agreement checks and batch audits

  • Domain specialists

    Medical, legal, financial and language expertise where the labels need it

  • Machine learning engineers

    Pre-labelling, active learning and export pipelines

Labels you can train on

Every project starts with a written labelling guideline and a pilot batch, labelled, reviewed and measured against your own expectations. The accuracy target, the review rate and the weekly volume are agreed from the pilot before the team grows, and each edge case the annotators meet is settled once and added to the guideline.

Your data stays under your control: annotators work in your labelling tool or in one set up in your cloud account, through named accounts whose access is removed when the work ends, never on files copied to personal devices. Personal data that the labels do not need is blurred or removed before labelling starts.

Other data and AI services we provide

Data and AI overview

Questions about data annotation

Which industries do your data annotators work in?

AI companies and model developers, for training, preference and evaluation data; automotive and robotics, for camera, LiDAR and sensor fusion datasets; healthcare, for medical images labelled under clinical review; e-commerce and retail, for product attributes, categories and search relevance; fintech and insurance, for documents and transactions; agriculture and manufacturing, for crop, field and defect images; and advertising and marketing technology, for content classification.

How do you measure annotation quality?

With numbers you can check. Gold items with known answers are hidden in every batch, a share of each annotator’s work is reviewed, agreement between annotators is measured on overlapping items, and each delivery comes with accuracy per label and a list of the guideline changes made.

Which annotation tools and formats do you work with?

CVAT, Label Studio, Encord, SuperAnnotate, Roboflow and Segments.ai, or your own in-house tool. Labels are delivered in COCO, YOLO, Pascal VOC, KITTI or JSONL, or in whatever schema your training pipeline expects.

How quickly can data annotators start?

When the right data annotator is available, the start is gated only by your interview and the NDA and IP assignment. Otherwise we run a search, which typically produces candidate profiles within two to three weeks, and nobody starts until you have said yes.

How do we hire data annotators through BigTree108?

Tell us the work, the seniority and the hours you need. We propose one or two people with their profiles, you interview them the way you would interview your own hire, and you sign one agreement with BIG TREE 108 LLC and receive one invoice a month.

Who owns the work they produce?

You do. Every specialist has a signed contract with BigTree108 that assigns all work product to the company, and our agreement with you assigns it onward. Code, designs and documents are delivered into your own repositories and tools, not kept where only we can change them.

Have data that needs labels?

Tell us the data type, the volume, the labels you need and the accuracy you expect. You get an answer within one business day: a plan for a pilot batch, a first draft of the labelling guideline, or candidate profiles.

Email hello@bigtree108.com How we respond