Automora

Prestigious car dealerships around the world rely on third parties to touchup images of their vechicles with long turnaround times of sometimes over 48hrs. My friend Ben came to me with this problem and we came up with Automora. Automora allows the editing to be done in-house, quickly, saving huge amounts of time and money by getting cars to market quickly.

Projects / Automora

We’ve reduced the average turnaround time from 24-48 hours to under 10 minutes by utilising machine learning models and computer vision techniques to perform all the changes and edits that would’ve been done using one of these third parties, but much quicker. Automora features background replacement, auto-centering & resizing of the subject, object & blemish detection and removal in the road surface, AI advert description generation, placeholder video creation & more.

Technical Architecture

The platform leverages a sophisticated distributed computing infrastructure:

• Auto-scaling GPU Instances: Dynamically scales GPU-accelerated compute resources based on Celery queue sizes, ensuring optimal resource utilization and cost efficiency
• Machine Learning Pipeline: Integrates multiple custom fine-tuned ML models across different stages of the processing pipeline, combining open-source models with proprietary optimizations for superior results
• Distributed Task Queue: Celery-based distributed computing for parallel processing of image and video transformations
• Enterprise Integrations: FTP integrations with major automotive companies for seamless data exchange and workflow automation
• Cloud Infrastructure: Built on AWS Fargate & EC2 with API Gateway for scalable, reliable service delivery

Automora is built mainly using Python, making use of libraries such as Celery for distributed computing and Flask for an internal API running on AWS Fargate & EC2. The system employs custom fine-tuned machine learning models throughout the processing pipeline, combined with open-source models for optimal performance. Enterprise FTP integrations enable seamless data exchange with major automotive companies. We are also working on a public facing API built using Amazon’s API Gateway and other similar technologies. If you need any assistance using cloud technologies or constructing any similar applications, don’t hesitate to get in touch.