Financial Services

Delivering Millions of AI-Powered Auto Repair Estimates for Better Claims Experiences


CCC uses NVIDIA DGX Cloud to manage and orchestrate development workflow, speeding the delivery of AI-based solutions.


CCC Intelligent Solutions Inc.

Use Case

AI Training


NVIDIA AI Enterprise, NVIDIA Base Command, NVIDIA DGX A100, NVIDIA DGX Cloud

Founded in 1980, CCC Intelligent Solutions (CCC) is a leading software-asa service (SaaS) platform for the property and casualty (P&C) insurance economy, powering operations for insurers, repairers, automakers, part suppliers, lenders, and more. More than 300 insurers and 28,000 collision repair facilities nationwide use CCC’s technology to process claims and repairs with speed and improved efficiency. CCC runs a 100 percent cloud production environment, providing customers with a flexible platform for continuous innovation.

CCC’s vision is to accelerate the digital transformation of the insurance and automotive industries by using AI-enabled solutions that automate workflows and enable better decision-making. The company is the first in the auto insurance claims industry to enable the straight-through processing of repair estimates. CCC® Estimate (STP) uses AI, insurer-driven rules, and CCC’s vast ecosystem to digitize the claims estimation process in seconds, improving customer experiences. Customers process more than 16 million claims using CCC solutions each year. Insurance customers have processed more than 14 million unique claims using CCC’s computer vision AI through 2022.

Challenge: Supporting Data Scientists and Engineers in Training AI Models to Deliver Solutions to Market Faster

CCC has a large team of data scientists committed to innovating and expediting the auto claims and repair experience for its customers. However, managing multiple teams and workflows across many innovations can pose a challenge. “To constantly innovate and develop intelligent solutions for auto claims processing, we need to establish an optimal organization with the right tool sets to deliver AI-based solutions effectively,” says Reza Rooholamini, chief scientific officer at CCC. Rooholamini states, “Our data scientists play a crucial role in creating new solutions and the ability to build models, experiment, and easily integrate these models into our AI workflows is key.”

CCC has built an end-to-end hybrid-cloud AI development and training pipeline. This infrastructure leverages over 150 NVIDIA A100 Tensor Core GPUs, including DGX Cloud and additional NVIDIA DGX systems on premises.

The CCC development teams are using DGX Cloud as an AI-training-as-aservice solution to supplement on-prem capacity, support supercomputing demand spikes, and accelerate AI development overall.

With 25–30 percent of its data scientists and engineering teams’ time dedicated to experimentation, coupled with massive datasets that are growing each day, CCC needed to enable a more scalable, multi-platform, hybrid multi-cloud for its development environment.

Image courtesy of CCC

  • CCC wanted to minimize low-value, high-volume, repetitive tasks in claims estimations.
  • Established an end-to-end pipeline for training AI models using a cloud-based infrastructure and additional on-premises components.
  • NVIDIA DGX™ Cloud ensures CCC can train models easily from anywhere and get 24/7 access to GPUs on demand.
  • Integrated NVIDIA Base Command™ Platform into their development pipeline for dataset management and orchestration, delivering a 2X speedup in running their data scientists’ experiments.
  • This AI pipeline has enabled CCC to unleash new innovations in the market, including their CCC Estimate-STP technology that provides line level claim estimates in seconds based on insurer rules.

NVIDIA DGX Cloud: Streamlined Developer Workflow With On-Demand Access to Powerful Computing

“Because we have a core AI development pipeline that integrates into other pipelines and plug-ins, it’s important for us to have a centralized platform to manage and orchestrate AI compute resources for our developers. This ensures different teams working with shared resources get the compute they need for their projects,” says Andrey Ptashnik, lead enterprise architect at CCC.

“Within DGX Cloud, we are using Base Command Platform’s open API and compatibility with Kubernetes API to easily integrate it into our internal model development pipeline, leveraging it as a scheduling and resource manager.”

CCC collects millions of unstructured and structured data points every year, so having the freedom to transfer data across our training resources is of utmost importance to ensure speed and flexibility. “In addition to being able to launch jobs and automate them much faster, Base Command Platform also makes data transfers a lot easier. CCC’s customers process millions of claims every year with CCC solutions, generating a lot of data. Data needs to be extracted from the production environment into sub-lakes, moved to the location where the model is being developed, and anonymized before we can experiment on it,” Ptashnik said. “We are able to easily extract datasets into Base Command Platform and manage them. When we schedule jobs on DGX Cloud, we can pick up those datasets from Base Command Platform and organize and align ourcompute resources accordingly.”

Reza Horrieh, senior manager of AI infrastructure and security, AI enablement, at CCC, explained, “With DGX Cloud, our team gets access to a powerful AI developer infrastructure providing a 24/7 on-demand, dedicated development cloud, with a great support team behind it.”

Image courtesy of CCC

Results: 2X Speedup in Running Experiments and 30-Fold Expedited Model Development

Using their AI pipeline, CCC launched CCC Estimate – STP in 2021, which can deliver a detailed line-level estimation of a collision based on insurer rules in seconds from just a few pictures from a mobile phone versus the several days that it can take using manual methods. This digitized estimation process has not only helped elevate the customer experience but is creating greater efficiency for customers.

“We develop an ensemble of big models and DGX Cloud enables us to train models and enhance features faster, while Base Command Platform enables us to ensure optimal utilization,” says Neda Hantehzadeh, PhD, director of data science at CCC. “ With Base Command Platform’s dataset management and orchestration capabilities, our data scientists have reported a 2X speedup in running experiments.”

DGX Cloud expedited model development and improved cost efficiencies. “With DGX Cloud integrated into our AI pipeline, we can expedite model development, developing some models in just a few days,” says Rooholamini.

Hantehzadeh added, “We run a lot of experiments on premises, but we may have spikes where we want to add, for example, 10 million more data points and do another run. If we don’t have enough capacity with our in-house systems, we switch to DGX Cloud. Base Command Platform makes this process seamless.”


  • 2X speedup in running data scientists’ experiments

  • Improved model development time from months to days

“Using DGX Cloud and Base Command Platform’s dataset management and orchestration capabilities, our data scientists have reported a 2X speedup in running experiments.”

Neda Hantehzadeh
PhD, Director of Data Science, CCCIS

Looking Forward

CCC plans to continue its investment in innovative and flexible technology to maintain its position at the leading edge of AI development. “The end-to end AI pipeline we’ve built is enabling us to unleash all kinds of innovations. We are injecting a lot of AI and automations into different products,” says Hantehzadeh. “Demand for training of models is increasing, and the ability to use cloud-native tools that easily access our powerful AI infrastructure, better manage our resources, and manage experimentation and development across hybrid clouds as needed will aid us in rolling out new solutions to the market quickly.”

Looking forward, the team expects to continue leveraging their AI pipeline integrated with DGX Cloud for multi-node training of large-scale computer vision and language models.

“We run a lot of experiments on premises, but we may have spikes where we want to add, for example, 10 million more data points and do another run. If we don’t have enough capacity in house, we switch to DGX Cloud. Base Command Platform makes this process seamless.”

Neda Hantehzadeh
PhD, Director of Data Science, CCCIS