Training data, labelled by a team of 2,200 in India.

    We have run data and document services since 2006. We now apply that scale to image, video, 3D, text and speech annotation, and to data collection for robots and other physical AI.

    2,200people on payroll
    Since 2006running data and document services
    ISO 27001certified. SOC 2 in progress
    13 languagesIndian regional languages, plus English

    Every self-driving car and robot starts with labelled data

    Cameras, LiDAR and wearable capture rigs produce millions of frames. People turn them into the examples a model learns from. That is the work we do.

    A self-driving car with roof-mounted sensors driving along a city street
    Driving dataRoad objects, lanes and signals labelled in 2D and 3D.
    A blue industrial robot arm in a factory
    Physical AITask video and sensor streams, collected and labelled.

    Send us a sample. Get a labelled pilot batch back.

    Share your guidelines and a few hundred items. We return labelled data with an accuracy report, so you can judge the work before you commit.

    Request a pilot

    Annotation services

    Work is done in your tool or ours, against your guidelines, with a reviewer on every batch.

    Image and video

    Boxes, polygons, segmentation, keypoints and tracking across frames.

    • 2D bounding boxes and semantic segmentation
    • Lane lines, traffic lights and road objects
    • Object tracking in video

    3D and sensor data

    LiDAR and multi-sensor labelling for driving and robotics.

    • 3D cuboids on point clouds, single frame and sequences
    • Camera and LiDAR fusion labelling
    • Point cloud segmentation

    Text and speech

    Transcription, classification and review in English and 13 Indian regional languages.

    • Transcription and speech labelling
    • Intent, sentiment and entity tagging
    • Response ranking and model output review

    Document and data processing

    The work we have done for BFSI, telecom, healthcare and government clients since 2006.

    • Data entry, conversion and digitisation
    • Data validation and verification
    • Loan and document processing

    Data for physical AI

    Robots and embodied models learn from what people do in the real world. We help teams collect it, clean it and label it.

    Collection

    Trained operators record everyday tasks in homes, workplaces and public spaces to your capture specification, with consent and privacy checks.

    Labelling

    Actions, objects, hand poses and task steps labelled on video and sensor streams, with review at every stage.

    Review of robot runs

    Success and failure tagging on recorded runs, so your team can find problems faster.

    How a project starts

    Pilot

    Send us your guidelines and a sample. We label a small batch and report accuracy and speed.

    Ramp

    We train and certify a dedicated team to your guidelines, then agree the SLA and review process.

    Scale

    Volume grows in phases. You get regular quality reports and one point of contact.

    Partner with us or work with us

    About 30% of our employees are persons with disabilities, and we have created work for more than 7,000 people who face barriers to employment. Our partners and freelancers extend that reach.

    Data partners

    We are looking for organisations that can supply or co-collect data, or send us annotation work as a subcontracting partner.

    • Licensed or custom datasets, with clear rights and consent
    • Language, speech and regional data
    • Platforms that need a reliable delivery team in India
    Become a data partner

    Freelancers and annotators

    We add trained freelancers when projects need extra hands or a specific language.

    • Paid per task or per hour, depending on the project
    • Training and guidelines provided before you start
    • Reliable internet and a laptop or desktop required
    Apply as a freelancer

    Questions we hear often

    How is our data kept secure?

    Data is encrypted with SSL/TLS in transit and AES-256 at rest. Access is role-based with multi-factor authentication. Vindhya is ISO 27001 certified and SOC 2 is in progress.

    Can we use our own annotation tool?

    Yes. We work in your platform or in a standard tool, whichever suits the project.

    Which languages can you cover?

    English and 13 Indian regional languages. Tell us the language in the enquiry form and we will confirm coverage.

    How do you price?

    Fixed monthly team pricing or per-transaction pricing, set after the pilot. Send your volumes in the form for a quote.

    Where do you work from?

    Our teams work across India, with delivery centres in multiple states.