Artificial intelligence (AI) has become the driving force behind the technology industry. It has propelled NVIDIA to over $2 trillion in market cap, turned Sam Altman into a tech demigod, and minted new AI unicorns. While the common theme among those developments is the hype around generative AI (GenAI) reaching the levels of insanity, there are startups pushing the boundaries in other areas of AI.
One such startup is Copenhagen-based FocalX, which is building AI and SaaS-driven inspection software for the automotive industry, and rethinking the impact of vision AI in process. Founded in 2021 and led by a dynamic team of Isabella Agdestein, Pritam Bolenwar, Nishchith Uchil, and Carlos Pasquali, FocalX aims to bring computer vision to the same level of prominence as GenAI.
Today, you can point your smartphone camera at a piece of clothing and instantly buy it online. Using AI and cloud-based technology, FocalX is bringing the same convenience to vehicle inspection, making it as simple as taking a photo of a vehicle and instantly diagnosing its condition.
The technology has the potential to completely transform vehicle inspection for pick-up and drop-off vehicles, off-road vehicles, and fleet management. Behind this novel technology solution lies a difficult journey for the founders, and their story is reminiscent of the trials and tribulations associated with an AI company.
Allure of entrepreneurship
The allure of entrepreneurship was one of the primary reasons Agdestein began building FocalX. She says this allure was especially “in areas that push the limits of current technology, and seemed like the right path for making a tangible impact.”
“When Pritam and I started building, our motivation was clear: to build a company that not only addressed current challenges but also anticipated future needs, staying at the forefront of innovation,” she explains.
It wasn’t just about building innovative products but also being part of developing the world we are living in and that journey came with its own set of challenges.
Agdestein, who leads FocalX as CEO, doesn’t shy away from acknowledging the challenge associated with building AI technology from the ground up, which is not only complex but also data-intensive. She says the combination of time spent experimenting and limitations associated with data sets was quite a hurdle.
The early validation of their solution from a major vendor allowed them to continue building it with the knowledge that they were “heading in the right direction, despite the challenges.”
Data challenge
Pasquali, VP of customer success at FocalX says their biggest hurdle came in the form of having large volumes of high-quality data to train their systems and build the models.
In the world of AI and large language models, it is a well-known fact that these systems could suffer from bias and hallucination if not trained using quality data.
For FocalX, it was a no-brainer to eliminate them at the start and one of the strategies they adopted from the beginning was to train their system with data that showcased “real-world situations.”
When it comes to quality data to train their AI model, Pasquali says they ensured that the vehicles were not pristine clean but instead were the ones on the streets, in different weather conditions, etc.
“We needed large data sets of these scenarios to be able to start building a computer vision technology that could provide value in real life,” he tells me.
For a young startup like FocalX, getting access to such a large data set with diverse yet real-time information is akin to climbing Mount Everest.
They overcame this challenge by partnering with two large global customers who were able to provide the startup with those datasets and keep generating them while they adopted FocalX’s product.
“It wasn’t easy,” says Pasquali, adding, “because there had to be a lot of expectation management when it came to AI at the start.”
Cloud and AI to Transform Vehicle Inspection
If you have seen a driverless vehicle then you already know that AI has already entered the world of the automotive industry.
Before we get to the day when every car is fully autonomous or the technology is mature enough to support such a transition, AI will make headway in other areas.
One area where the impact of AI could be immediate is vehicle inspection and FocalX is at the leading edge of it.
The Danish startup has built an AI-powered cloud-based vehicle inspection platform that works through a combination of user-friendly applications and APIs.
The customers of FocalX have the option to either follow the app user flow to create an inspection process or they can create their user flows and plug into FocalX through API.
Once the vehicle images are uploaded to the backend of FocalX, they are then processed using a machine learning engine.
“We have several models that are simultaneously working on the image,” says co-founder and CTO Bolenwar. He adds, “The AI damage results are promptly sent back to users, who can access their inspection reports as PDFs, in our web system, or through the app.”
The platform also integrates with the customer backend to send these reports directly via API. While reiterating the challenge of building the AI models for their computer vision product, Bolenwar says the lack of data restricted them from experimenting with different versions of their model and seeing how they improved the results. He says they also had to be cognizant of the fact that the literature in the field of computer vision is doubling in no time.
“To be at the forefront of technology requires a lot of research into new methods, and conducting experiments based on the latest research. This is an ongoing and fairly time-consuming effort, yet extremely important,” he explains.
Despite the challenges, FocalX has built several different AI models which run on every image uploaded to their system. Their initial stage models detect things such as whether what they are seeing is a car, the position of it and car parts present in that image.
Once they have verified the vehicle, its position, and car parts present in the image, FocalX runs all of its damage models followed by validation models for the detected damages and finally, a severity model, which defines the severity of the damages.
AWS to bolster the tech
To support its cutting-edge tech, FocalX needs an efficient, scalable, and compatible cloud service. Bolenwar says they found all those attributes in Amazon Web Services (AWS) and they chose the service because of “AWS Aurora’s robust PostgreSQL-compatible database management, and AWS SageMaker’s capabilities for AI and ML model development.”
He adds that the cost-effectiveness of AWS’s pricing model and the secure, accessible data storage offered by AWS S3 buckets were crucial factors too. Like many entrepreneurs, Bolenwar and his co-founders had prior experience with AWS, which made it easier for them to seamlessly integrate its internal tools, and further streamline the development and deployment.
“AWS significantly bolsters a tech-oriented startup like FocalX by providing a scalable and secure system. Services such as AWS Lambda, mailing systems, SQS, and serverless databases enable efficient backend management with fewer resources,” he says.
While reliability and scalability played a key role in FocalX choosing AWS, they also acknowledge that credits offered periodically help in experimentation and innovation. “AWS aids in expanding our reach and scaling faster by providing a scalable system that enables us to handle a large number of requests,” Bolenwar adds.
With its multi-region support and CDN services to serve customers better, AWS further adds strength to FocalX’s business operations. The startup is even considering using AWS Marketplace from early 2025 to further enhance the capabilities of its platform. With AWS Marketplace, FocalX will gain access to a wide range of software and services that can further enhance the capabilities of its platform.
Efficiency and Adoption
Nearly three years after its launch, FocalX continues its focus on tech and product development while also looking to expand its commercial team to better meet the demands of its customer base. The startup, which has more than 20 people working on the tech side alone, is further looking to build its team as the battle rages around the world for the smartest and brightest minds in the world of AI and ML.
“We’re constantly looking for the best and brightest in AI & ML who can introduce innovative research and ideas previously unexplored by our team,” says Uchil, co-founder & VP of Product.
With GenAI gaining widespread attention, the startup sees an AI world where computer vision solutions like FocalX could become an integral component of most software systems. These systems will be tasked with analysing the current state and automating routine tasks, the tasks that humans spend hundreds of unproductive hours doing.
Bolenwar goes as far as to say that this might not have the same wow factor as GenAI but they see a heydey for such a computer vision system. Till then, Uchil says FocalX will continue to focus on efficiency and adoption, the two key factors on which its immediate success depends.
This vision is also supported by some of Denmark’s most successful entrepreneurs, who backed FocalX’s early funding round. It has also received several grants, which Agdestein sees as an acknowledgement of the novelty and promise of its solution. She adds the startup is preparing to enter the market again for funding to help solve the challenges of many automotive businesses.
One thing that remains extremely important, Uchil adds, is the rapid adoption of its solution within the automotive industry. By demonstrating their ease of integration into existing workflows and immediate benefits, FocalX wants to bring its AI to a stage where it significantly boosts clients’ operational efficiency.
Looking for more inspiring stories like this? Join AWS Summit Stockholm to connect with industry leaders and top regional startups, learn where investors are placing their bets and immerse yourself in technical brilliance. Register now to elevate your startup journey!
01
Meet the 10 promising startups selected for Glovo Startup Lab