In a world where artificial intelligence is a key driver of innovation, the speed and reliability of infrastructure determines the success of a project. A team of developers from Denmark faced the need to deploy a platform for building an analytics system for a supermarket retail chain using the YOLO neural network.
The project had strong practical business relevance, and the INTROSERV team immediately began selecting and setting up the platform according to the client’s technical specifications.
Context: How YOLO and AI Improve Supermarket Operations
A development team specializing in AI solutions for retail approached INTROSERV with a request for an inference environment for a computer vision neural network - Ultralytics YOLOv11.
Their goal: Analyze customer behavior in a small supermarket chain to increase average order value, assess the effectiveness of promo stands and banners, optimize checkout performance during peak hours, determine those peak hours per location, and optimize product placement based on hot and cold zones within the store.
Management realized that existing analysis methods (based on sales reports and managers’ intuition) were insufficient. A deeper, technology-driven understanding of customer behavior was needed.
This AI analytics project was planned to be rolled out across all key branches, starting with several newly opened locations.
This project demanded more than just "renting a server" - it needed a specialized, secure environment ready to handle specific AI workloads.
The Challenge: How to Collect Accurate Customer Behavior Data Using AI
The core challenge: build an actionable, efficient analytics platform for retail where simple statistics fall short.
How to capture accurate data on the movements of hundreds of customers without disrupting their shopping experience? And how to turn that data into concrete, measurable business insights?
While the AI development team tackled the neural network integration, the infrastructure setup was fully entrusted to INTROSERV.
The project immediately piqued the interest of INTROSERV’s technical specialists. Thanks to a clear set of goals and requirements, a suitable configuration was already in stock and ready for rapid deployment. The INTROSERV team knew exactly how to securely transfer multiple high-resolution video streams to the data center.
The Solution: Dedicated GPU Servers with Proxmox for YOLO-Based Analytics
The INTROSERV team developed and deployed a multi-component solution using two dedicated servers with Proxmox VE and Proxmox Backup Server.
This allows for efficient resource separation, task isolation, and full VM backup for disaster recovery.
Infrastructure Overview: Multiple Virtual Machines
Virtual Machine #1: Network Gateway with pfSense
- Tasks: Secure connectivity and access control to server infrastructure.
- Implementation: A pfSense router and WireGuard VPN server were deployed. The supermarket’s internal network - where the video surveillance cameras operate - is securely connected to the data center server via an encrypted VPN tunnel.
Virtual Machine #2: AI Core (YOLO + GPU)
- Tasks: Fine-tuning the YOLO model with reference video frames and real-time object detection.
- Implementation: The key VM. Using PCI passthrough in Proxmox VE, the NVIDIA RTX 2080 Ti GPU was assigned directly to the VM. This provided near-native performance for:
- Model fine-tuning: Custom YOLO training on surveillance footage to increase detection accuracy.
- Recognition: Real-time video stream processing, human detection, and movement tracking.
Virtual Machine #3: Analytics & CRM
- Tasks: Collect, store, and visualize data.
- Implementation: A CRM system was deployed here to receive processed data from VM #2: person coordinates, dwell times, queue sizes. The system generates movement maps, heatmaps, foot traffic graphs, and detailed business reports.
Proxmox Backup Server:
- Tasks: VM backup repository.
- Implementation: Installed on a separate server, connected via a dedicated local interface to the main server. Proxmox Backup Server ensured automated and secure backups of all VM data.
Dedicated Server Configurations Used
YOLO server
- Platform: Asus ESC4000 G4
- Location: Netherlands (Naaldwijk)
- CPU: 2x Intel Xeon Silver 4208
- 16 Core 32 Threads 2.1/3.0GHz
- Memory: 64GB REG ECC DDR4
- Disk: 2x 500GB SSD Hardware RAID 1
- Disk: 2x 3.84TB NVMe Software RAID 1
- GPU: Gigabyte GeForce RTX 2080 Ti 11GB
- Port/Traffic: 2 Gbit/s - Unmetered
- VLAN: 10 Gbps VLAN
- AntiDDoS: DDOS 20Gbit/s protection
- Operating system: Proxmox VE 8
PBS
- Platform: Dell PowerEdge R240
- Location: Netherlands (Naaldwijk)
- CPU: Intel Xeon E-2236
- 6 Core 12 Threads 3.4/4.8GHz
- Memory: 16GB ECC DDR4
- Disk: 2x 250GB SSD Hardware RAID 1
- Disk: 4x 16TB SATA Hardware RAID 10
- Port/Traffic: 1 Gbps - 50TB Traffic
- VLAN: 10 Gbps VLAN
- IP-KVM: iDRAC
- AntiDDoS: DDOS 20Gbit/s protection
- Operating system: Proxmox Backup Server
INTROSERV: Your Technology Partner for Innovation
By outsourcing the entire infrastructure setup to INTROSERV, the client’s development team was able to focus fully on enhancing the AI model and working with data.
Instead of spending time on hardware selection, protection setup, and high availability configuration, the client received a production-ready platform - easily replicable across all future locations.
At the time of publication, the client has successfully integrated the analytics system, and the first retail location is already receiving daily customer behavior reports. These insights are helping management make informed decisions on sales optimization and product placement strategies.
This project proves that INTROSERV is not just a provider, but a true technology partner—ready to support complex IT projects with reliable infrastructure.
This proves that INTROSERV is not just a contractor but a true technological partner capable of ensuring success in complex IT projects.
To select the optimal infrastructure for your project, please contact INTROSERV's sales department.
