Anurag Jha
← All products

Spyne AI · 2021 to 2022 · B2B · Automotive, computer vision

AI imaging APIs for cars

Studio quality car photos without a studio: background replacement, shadows and reflections, window and body correction, sold as APIs to dealers and marketplaces in five regions.

My roleOwned the automotive vertical: vision, roadmap and life cycle. Worked with 6 full stack and 3 AI developers, a designer and 10+ annotators.

~$2.4MARR from the line I owned: ~80% of Spyne's ~$3MCEO interview, Mar 2023; my arithmetic
~10xARR growth yearCEO interview, Mar 2023
40% → 90%model accuracy in a yearMy figure
Background replacement, before and after.

The challenge

Dealers and marketplaces needed studio quality photos at catalogue scale.

What shipped

Segmentation, detection and generative pipelines (UNet, YOLO, GAN), colour correction, a guided capture app and 360 spins.

What moved

Model accuracy rose from 40% to 90%, and the automotive line was about 80% of revenue as ARR grew about 10x to ~$3M

A

Capture to catalogue

A phone, a guided capture, and an API that does the studio.

  1. 01

    Guided capture

    A shot checklist and 360 capture on the phone.

    Shot checklist.
    Android flow with 360.
    01 / 02
  2. 02

    The studio, as an API

    Background, shadows, reflections and colour, consistent across a whole inventory.

    Before and after.
  3. 03

    The data loop

    Weekly cycles on training data and annotation took accuracy from 40% to 90%.

    Working notes on the imaging stack.

Proof

~$3M ARR and 140 customers, CEO interview, Mar 2023.
$7M round led by Accel, Apr 2022, during my time.
Car backgrounding is still a live Spyne product.
01 / 03