Sequans Communications Enhances QA with an ML-Powered SBA

Business Challenge

Sequans Communications manages a large-scale, automated test environment to validate the behavior of its chipsets across complex telecom scenarios. To identify faults, performance issues, or anomalies, the company has a test log analysis procedure, which requires significant time when done manually. The logs, being the primary source of data, can be unstructured and complex to analyze. Besides, missing subtle deviations can impact end-product quality and time-to-market.

To ensure higher accuracy, faster fault resolution, and enhanced test cycle efficiency, Sequans Communications sought a smart, automated log analysis solution capable of identifying deviations from normal behavior across multiple campaign scenarios.

Solution & Business Value

Tietoevry Create implemented a System Behavior Analysis (SBA) solution for Sequans Communications based on our in-house machine learning research. We introduced a new temporal and sequential learner, capable of understanding log context and pinpointing anomalies with minimal false positives. The approach trains ML models using logs from successful test runs to establish a baseline of normal behavior. Using this baseline, the model identifies both deviations and their exact timing, allowing faster debugging and improved precision.

With minimal manual tuning, SBA can parse complex logs, detect patterns, and significantly cut down analysis time – from hours or days to minutes – helping experts to focus on important tasks and improving product quality.

Key Benefits:

• Time savings: Log analysis time (including training and parsing) reduced from hours/days to minutes, saving engineers’ time.
• Higher accuracy: The solution detects anomalies previously missed.
• Increased product quality: More reliable fault detection ensures better-performing chipsets.
• Secure by design: On-premises deployment ensures that no sensitive data leaves the client's infrastructure.
• Scalability: Initially integrated with 3 test campaigns, expanding to 60.

"The solution delivered by Tietoevry Create engineers gave us insights we couldn't see before. It accelerated our fault detection and showed real potential from the very first proof of concept. Now, in the next phases of SBA implementation, we are planning to integrate it into other test campaigns to maximize the impact across the validation process. We see this as a strategic step toward faster releases and higher product quality."

– Xianbo Meng, Product Validation Manager at Sequans Communications

Technical Details

The project began with a 4-week proof of concept, using Sequans Communications’ logs to validate solution accuracy. Next, it evolved into a multi-phase engagement:

• Phase 1: As an extension of our initial PoC, we conducted the initial analysis across three log sources and over 200 KPIs. Sequans Communications immediately observed faster and accurate fault detection and the ability to uncover hidden issues.
• Phase 2 (currently in progress): Focus is made on integration of the solution into Sequans Communications’ environment.
• Coming up next: SBA will be scaled across more test campaigns (up to 60).

The solution will also support sensitive data handling by operating fully on-premises, ensuring data security and compliance.

Tools & Technologies

• Python
• Machine Learning
• Angular (for frontend)
• Typescript (for frontend)
• Anomaly detection
• KPI analysis
• Log analysis