Salt Security: Detecting Threats, Anomalies & Vulnerabilities in API Traffic with AI
In a dynamic intersection of data science and cybersecurity, Salt Security faced a compelling challenge: How could their expanding data science team attain self-reliance and lessen its dependency on DevOps and engineering squads?
Their objective was clear — to amplify model delivery to production without burdening the engineering team further. This case study delves into how Qwak was enlisted to transform this challenge into an opportunity, ensuring both efficiency and autonomy for Salt Security's operations.
Salt Security delivers end-to-end protection for APIs throughout build, deploy, and runtime phases by combining comprehensive coverage with an ML/AI-driven big data engine. The platform not only stops attackers in the early stages of an attempted attack but also enhances API security posture by discovering all APIs, preventing unauthorized access, and offering remediation insights for development teams.
Industry
Cyber Security, Software/Technology
Use Case
Model Frameworks

As our data science team and customer base grew, we found it challenging to move our new and more sophisticated models into production, even though we knew the updated ones were better. We couldn’t stay in that mode.
Model Monitoring
By Design
Engineering Dependency
Challenges
With its accelerated growth, Salt Security expanded its Data Science team, leading to the generation of advanced ML models. However, transitioning these models to production presented multiple challenges:
- The AWS SageMaker framework demanded consistent involvement from the Infrastructure team for each new model deployment.
- Preliminary PoC experimentation often necessitated the expertise of a seasoned engineer.
- The data science teams, skilled in their field, lacked the engineering experience crucial for deployment.
- Integrating with Kafka's event-based architecture was intricate, especially when endpoints needed to manage and output prediction streams.
Solutions
With its accelerated growth, Salt Security expanded its Data Science team, leading to the generation of advanced ML models. However, transitioning these models to production presented multiple challenges:
- The AWS SageMaker framework demanded consistent involvement from the Infrastructure team for each new model deployment.
- Preliminary PoC experimentation often necessitated the expertise of a seasoned engineer.
- The data science teams, skilled in their field, lacked the engineering experience crucial for deployment.
- Integrating with Kafka's event-based architecture was intricate, especially when endpoints needed to manage and output prediction streams.