By switching to an end-to-end Computer Vision platform, SGS transformed its fragmented projects into a streamlined workflow, enabling a small team of four to manage over ten complex projects at the same time. This strategic shift not only accelerated their AI development - it also unlocked international scalability, reinforcing their position as an industry innovator.
- Time-to-Model Reduction: SGS slashed the time required to fine-tune a computer vision model for a new lab from 3 days to just 1 day.
- Team Efficiency: A small team of just two data scientists and one front-end developer now efficiently manages several computer vision projects in parallel, from development to production.
- Global Scalability: Overcame critical tooling roadblocks to unlock international expansion, deploying proven AI solutions to labs across Europe.
About SGS
SGS is the global leader in testing, inspection, and certification, with 100,000 employees dedicated to enabling a better, safer, and more interconnected world. The company works with clients across nearly every industry to ensure products and processes meet quality, safety, and compliance standards. SGS relies on an extensive network of laboratories specializing in a wide range of fields, including food, consumer goods, pharmaceuticals, petroleum products, environmental waste analysis etc…
SGS's CV Journey: The Challenge of Scaling In-House AI
SGS's journey began with a clear goal: use AI to digitize and accelerate manual lab analyses, like counting bacteria in petri dishes or measuring the dimensions of food products. The first attempt involved outsourcing AI development, but this approach quickly proved to be a "black box". Expensive, inflexible, and not precise enough for their specialized needs.
Taking development in-house, the team built a workflow using a collection of disconnected tools. It was functional for initial proofs-of-concept, but the fragmented system created big bottlenecks as they tried to scale. It was difficult to maintain, lacked traceability, and made collaboration with lab experts across different locations nearly impossible.
"Picsellia unified our entire workflow. We went from wrestling with a fragmented set of tools to having a streamlined process, which let us focus on what actually matters: building high-quality models and scaling our impact across the business."
The Solution: A Unified Platform to Industrialize Computer Vision
SGS recognized that to scale effectively, they needed to move beyond their home-made solution. They were looking for a professional platform that could streamline their entire AI lifecycle, from data management to production monitoring.
With Picsellia, they found the perfect fit, offering an end-to-end platform that replaced their entire fragmented toolchain. It gave them a centralized hub for data annotation, model training, and experiment tracking, solving their core challenges around traceability, reproducibility, and collaboration. Most importantly, it provided the robust, scalable foundation they needed to industrialize their computer vision projects.
Business Impact of Computer Vision with Picsellia
By centralizing its CVOps on Picsellia, SGS achieved significant and measurable improvements across its operations.
- Drastically Reduced Development Time: The efficiency gains were immediate and significant. By using Picsellia’s streamlined workflows, the team reduced the time to fine-tune and adapt a model for a new lab from three days down to just one. This 66% reduction allows them to respond to internal needs quicker and more accurately.
- Massive Gains in Team Efficiency: Picsellia acts as an enabler for the SGS data science team. A small team of four is now able to successfully develop, deploy, and manage over ten distinct computer vision projects. The ability to easily duplicate project structures and reuse assets means they can launch new applications without starting from scratch.
- Unlocked International Expansion: Before Picsellia, SGS was unable to expand its AI solutions internationally due to the lack of a unified platform for collaboration and annotation. Picsellia removed this barrier entirely. Its collaborative features were the key enabler that allowed SGS to deploy their mature solutions to labs across Europe, turning a local success into a global capability.
- Enhanced Quality and Traceability: In a compliance-driven industry, reproducibility is very important. Picsellia provides a single source of truth for all datasets, experiments, and models. This ensures full traceability, eliminating the risk of losing work and empowering the team to build robust, reliable, and auditable AI solutions.
Moving Forward: With a scalable and powerful CVOps foundation in place, SGS is ready to accelerate their AI innovation even further. The team plans to continue the international rollout of its mature computer vision solutions while also tackling new applications for other SGS divisions in France, including complex challenges in ecotoxicology, cosmetic and environmental sectors. Picsellia provides the collaborative environment needed to ensure SGS remains at the cutting edge of technology-driven quality assurance.
