From Damage Photo to Instant Quote: How DingGo Took AI to Production on AWS with base2Services
More recently base2Services have been the ones pushing us on better infrastructure for a longer term AI strategy, from pulling AWS credits to run proof of concepts to onshoring our PII workloads through Amazon Bedrock. We now run separate scopes for our core product and our AI. That strategy, modelling and management is a skill set no one in our business wants to own, and they are a great partner for it. We would be investing a lot more to have that level of expertise in-house.
Josh Sandford, Co-Founder and Chief Product Officer (CPO), DingGo
See the impressive results
- Took DingGo's Crash Intelligence AI from experiment to production on AWS, turning a damage photo into an instant repair quote
- Fine-tuned a proprietary computer vision model on a dataset that has grown beyond 700,000 labelled damage images, with human-in-the-loop corrections feeding continuous retraining
- Built a two-track AI architecture that keeps PII workloads onshore in Amazon Bedrock while routing non-sensitive workloads across multiple models for cost and flexibility
- Used AWS credits to run multi-model proof of concepts, letting DingGo choose models on evidence and de-risk the investment
- Separated AWS scopes for the core platform and the AI workloads to prevent contention and keep spend controlled
- Wrapped the program in guardrails, human oversight and governance ready for the AI assurance standards arriving in Australia
- Supports DingGo's board level goal to reduce cost to serve by 50 percent through AI driven automation
- All on a 24/7/365 managed AWS foundation built over a partnership of more than five years
A bit about DingGo
DingGo is a market-leading digital car crash repair marketplace redefining fleet accident and repair management. Managing over 75,000 vehicles across Australia and New Zealand, DingGo eliminates outdated and manual processes by connecting drivers, repairers, insurers and fleet managers in a single, intelligent ecosystem.
Designed to disrupt a traditionally fragmented industry, DingGo's platform delivers end-to-end crash management, using automation, real-time data and one of the largest independent repair networks to provide speed, transparency and control at every stage.
What DingGo needed
DingGo had already proven its platform at scale. After years of rapid growth the focus shifted to efficiency and differentiation, with a board level goal to reduce cost to serve by 50 percent. AI was the lever, but DingGo wanted to use it deliberately rather than chase headlines.
The opportunity sat in DingGo's own data. With hundreds of thousands of damage photos and matching repair quotes, the company could build models its competitors could not. The challenge was production, not ideas. DingGo needed a secure way to fine-tune and serve models, somewhere to keep personally identifiable information onshore and under its own control, a way to test models without runaway cost, plus the guardrails and human oversight to use AI responsibly.
DingGo runs a deliberately lean team with no DevOps or AI infrastructure function of its own. It needed a partner who could provide that engineering depth, push the infrastructure forward and let DingGo's people stay focused on product and customers.
How base2Services delivered
Through our AI Factory, base2Services gave DingGo the infrastructure, MLOps practices and governance to move its AI ambition from experiment to production, without DingGo having to build a deep AI engineering team in-house.
Turning proprietary data into a production model
DingGo's goal was simple to describe and hard to deliver: take a photo of vehicle damage and return an accurate repair quote in seconds. Achieving it meant chaining several models together to detect the damage, classify its type and severity, identify the vehicle and match it to a cost estimate. We supported DingGo's Crash Intelligence initiative with AI and machine learning infrastructure on AWS, building pipelines to retrain the model as new data arrives and securing the environment it runs in. The trained computer vision model was deployed to Amazon ECS Fargate and exposed through Amazon API Gateway so it could serve the live platform. As DingGo's operators triage and correct quotes in their normal workflow, that human input becomes training data that continuously improves the model, a human-in-the-loop pattern that has taken the dataset from an initial sample to more than 700,000 labelled images.
A two-track architecture for sensitive data
Not all data can be treated the same way. DingGo handles personally identifiable information that has to stay onshore and under its own control, alongside non-sensitive data such as damage photos and panel quotes. We helped DingGo separate the two. Sensitive workloads run on locally hosted models within DingGo's own AWS account using Amazon Bedrock, so regulated data never leaves their control. Non-sensitive workloads run through a flexible, multi-model approach that lets DingGo choose the best model for each task and switch as better options appear. Guardrails, access controls and human oversight sit across both tracks, giving DingGo a foundation ready for the AI assurance standards now arriving in Australia.
De-risking experiments and controlling cost
Experimenting with AI can get expensive quickly. To keep DingGo's investment focused, base2Services arranged AWS credits so the team could run proof of concept comparisons across open source, commercial and their own earlier models without large upfront cost, then choose on evidence rather than hype. We also separated DingGo's AWS scopes for the core platform and the AI workloads, preventing contention and keeping spend visible and controlled. All of it runs on the 24/7/365 managed AWS foundation we have operated for DingGo across a partnership spanning more than five years.
The result is an AI program that delivers real business value. Rather than chasing the newest model for its own sake, DingGo can act on proprietary data with confidence, keep regulated information safe and point AI at the work that moves its cost to serve. As their DevOps and AI Factory partner, base2Services continues to manage, secure and evolve the platform as DingGo's strategy grows.