Just Got Funded.
← All companies

CVAT just raised $2.5M.

Teaching AI what it's looking at

AIOpen source

The round

Raised
$2.5M
Filed
Oct 28, 2026
Based in
Wilmington, DE
Investors
2 in this round
Legal name
CVAT.ai Corp
Website
cvat.ai

About CVAT

The problem

Before an AI can spot a stop sign or a broken bone, people must label thousands of examples by hand.

The usual way

People outline objects one image at a time, for days or weeks.

What they built

CVAT is a free, open-source labeling tool with AI helpers built in, spun out of Intel and downloaded millions of times.

CVATOpen source
LabelsImages, video, 3D, audioAll in one
HelpersAI tools pre-labelBuilt in
Team300+ human labelersOptional

How it works

  1. UploadImages, video, 3D scans or audio
  2. DraftAI tools suggest the labels
  3. ReviewPeople check, then train the model

Traction

2.8M+downloads on Docker HubPer cvat.ai

Used by

  • Deloitte
  • Purdue
  • EPAM

Their website

cvat.ai
CVAT website
cvat.ai, captured for this episode.

The script

Every scene of the video: the narration, what's on screen, and a still from the final cut.

  1. Scene 1: Hook
    01Hook0-3s
    “CVAT just raised two and a half million dollars.”
    On screen

    $2.5M · Teaching AI what it's looking at

  2. Scene 2: The problem
    02The problem3-9s
    “Before an AI can spot a stop sign or a broken bone, people must label thousands of examples.”
    On screen

    AI learns from labeled examples.

    Self-driving cars and X-ray AI need thousands of images tagged by people first.

  3. Scene 3: The usual way
    03The usual way9-13s
    “That means outlining objects by hand, one image at a time.”
    On screen

    Box by box, by hand.

    People outline objects one image at a time, for days or weeks.

  4. Scene 4: The product
    04The product13-18s
    “CVAT is a free, open-source tool for labeling, with AI helpers built in.”
    On screen

    Meet CVAT, AI's labeling tool.

    • Labels Images, video, 3D, audio
    • Helpers AI tools pre-label
    • Team 300+ human labelers
  5. Scene 5: How it works
    05How it works18-25s
    “Upload your images or video. Let AI draft the labels. Then people check them, and the data is ready.”
    On screen

    How it works.

    1. Upload Images, video, 3D scans or audio
    2. Draft AI tools suggest the labels
    3. Review People check, then train the model
  6. Scene 6: Proof
    06Proof25-30s
    “It started at Intel, and it's been downloaded more than two point eight million times.”
    On screen

    Born at Intel, used everywhere.

    2.8M+ downloads on Docker Hub

  7. Scene 7: The round
    07The round30-34s
    “That's two and a half million dollars, from two investors.”
    On screen

    Better labels, better AI.

    Wilmington, DE · 2 investors · 15K+ GitHub stars

  8. Scene 8: Outro
    08Outro34-37s
    “Follow us for a new funded startup every day.”
    On screen

    Tomorrow: another startup that just got funded.

Caption, as posted

AI startup CVAT just raised $2.5M to help teams label the images and video that AI learns from.
Follow @just.got.funded for a new funded startup every day.

Before an AI can spot a stop sign or a broken bone, people must label thousands of examples by hand.
CVAT is a free, open-source labeling tool with AI helpers built in, spun out of Intel and downloaded millions of times.

Did you know AI needs humans to label its homework first? 👇
Send this to someone who's building with AI.

#artificialintelligence #computervision #opensource #delawarestartups #justgotfunded

Sources