Trained beforeyou hire.Ready on day one.

A&L Factory builds a bench of technical annotation and labelling professionals. Everyone completes training on open-source tools and passes an assessment before joining your project, for any industry and any type of data.

Hire trained talent Join the talent pool
  • Trained on open-source labelling tools
  • Assessed against gold-standard labels
  • Profiles for every industry and data type
street_0412.jpg
    1. Labelling
    2. In review
    3. Approved

    Annotation and labelling for every kind of data

    Our teams label the data your models learn from, whether that is a single image class or a multi-sensor pipeline.

    image

    Image annotation

    Precise outlines and labels for computer vision models, from simple classes to pixel-level masks.

    • Bounding boxes and polygons
    • Semantic and instance segmentation
    • Keypoints and landmarks
    • Image classification and attributes
    video

    Video annotation

    Objects and events followed across frames.

    • Object tracking with interpolation
    • Frame-by-frame labelling
    • Action and event tagging
    • Temporal segmentation
    text

    Text and NLP

    Structured labels on language data.

    • Named entity recognition
    • Sentiment and intent
    • Text classification
    • Relation extraction
    audio

    Audio and speech

    Recordings segmented, transcribed and tagged.

    • Transcription
    • Speaker labelling
    • Sound-event tagging
    • Emotion and intent labels
    3d_geo

    3D, LiDAR and geospatial

    Depth and map data labelled in three dimensions.

    • Cuboids on point clouds
    • Sensor-fusion labelling
    • Satellite and aerial imagery
    • Land-cover mapping
    documents

    Documents and OCR

    Forms, invoices and scans turned into data.

    • Text region marking
    • Key-value extraction
    • Tables and line items
    • Handwriting review
    genai_llm

    GenAI and LLM data

    Human feedback that improves model responses.

    • Response ranking and preferences
    • Prompt and answer writing
    • Factuality and safety review
    • Evaluation sets
    review_qa

    Review and quality assurance

    A second pair of eyes on labels you already have.

    • Audits of existing datasets
    • Gold-set checks
    • Relabelling and clean-up
    • Guideline and ontology support

    How we keep labels accurate

    Clear guidelines

    Every project starts with written label definitions, examples and edge-case rules.

    Gold-standard tests

    Annotators are scored against pre-labelled sets before and during the project.

    Agreement checks

    Overlapping tasks show how consistently different people label the same data.

    Layered review

    Reviewers audit samples, send corrections back and track the types of error.

    Built for every industry

    Pick an industry to see what usually gets labelled. Guidelines and label classes are set up around your project, so treat this as a starting point.

    What gets labelled

    Typical label classes

    Request a team

    Your industry is not listed? We write a custom guideline and label set, then train the team on it.

    Training is finished before the project starts

    Each professional in the pool follows the same programme, then specialises. Nobody is placed on a project until they have passed the assessment.

    1. 1

      Foundations

      What labels are for, how to read guidelines and label sets, handling edge cases, measuring accuracy and protecting sensitive data.

    2. 2

      Tool labs

      Hands-on practice in the open-source tools listed below, including shortcuts, tracking, interpolation and model-assisted pre-labelling.

    3. 3

      Domain track

      Choose an industry, such as vehicles, medical, documents, speech or retail, and practise on reviewed sample sets.

    4. 4

      Assessment

      A timed test against gold-standard labels. Accuracy, speed and agreement with reviewers are all scored.

    5. 5

      Bench-ready

      Trainees who pass receive a profile, are matched to project types and can start when you need them.

    Every trainee learns

    • Reading and applying labelling guidelines
    • Label classes, attributes and edge cases
    • Quality measures such as precision, recall, IoU and inter-annotator agreement
    • Export formats including COCO, YOLO, Pascal VOC, JSON and CSV
    • Data privacy, confidentiality and secure working habits

    The technical track adds

    • Self-hosting CVAT and Label Studio with Docker
    • Python scripts that convert formats and validate labels
    • Model-assisted pre-labelling with models such as Segment Anything and YOLO
    • Writing labelling configurations and custom interfaces
    • Reading model errors to find gaps in the labels

    Open-source tools we train on

    Training uses open-source tools, so skills carry across projects and your team is not tied to a paid licence.

    Open-source tools have no per-seat licence fees. We can train on the tool your project already uses, and self-host CVAT or Label Studio for you.

    Profiles ready to join

    Every profile below has completed training and assessment. Add the ones you need to a request and we will come back with matching people.

    Build your team, or join one.

    The bench only works if trained people and real projects meet. Tell us which side you are on.

    For companies

    Tell us the data type, industry and profiles you need. We reply with the trained people who fit and a plan to start.

    For talent

    Apply, complete the training, pass the assessment and join the bench. Companies hire from the pool as projects begin.

    Data types
    Profiles needed

      None chosen yet. Pick profiles above or describe what you need below.

      Continue to request