The Path to Success with MYRMEX

RESULTS WITH GOOGLE CLOUD

• ✓ About 73% improvement in processing time for a final response in complex actions
• ✓ Reduction in response time for basic interactions, from 2 minutes to 30 seconds
• ✓ 45% speed gain in deploying new AI agents
• ✓ Model lifecycle accelerated by over 35%
• ✓ Operational excellence through GKE Autopilot, eliminating Kubernetes cluster management complexity

Upon migrating to Google Cloud, the company used cloud solutions to implement significant improvements in its platform performance, directing 100% of its engineering efforts to product innovation

AINEXT is an innovative technology startup focused on solving a critical market problem: the operational complexity and fragmentation of security tools. The company developed Myrmex, a cloud-native multi-agent AI platform to audit, operate, and document complex infrastructures.

Myrmex distinguishes itself with its unified and autonomous approach, primarily focused on prevention, interacting with the client's environment to audit and analyze risks or configuration failures, apply fixes, and assist technical operations.

In an initial phase, AINEXT faced significant challenges with its previous technology service provider. The main obstacle was operational inefficiency caused by infrastructure management complexity, along with situations that impacted the development speed of the company's solution, consuming precious resources.

The search for a new partnership that could overcome these problems led AINEXT to Google Cloud, solidified after the company's contact with the cloud technical team at Startup Summit 2024 - an event exploring advances in technology, artificial intelligence, and digital transformation.

"The decision to migrate was made specifically aiming at the use of tools such as Vertex AI and the Agent Development Kit (ADK). The automated infrastructure enabled by GKE Autopilot, scalability, and the possibility of our business growth were also key factors."

Pedro Verissimo — Head of Cloud, AINEXT

New Journey and the Triad: Vertex AI, ADK, and GKE in Action

AINEXT's service provider migration project began days after the first interactions with the Google Cloud team, between October and November 2024. Thanks to the Myrmex platform architecture, which was already based on containerized microservices, the process was remarkably fluid.

The transition, involving the complete migration of all workloads to Google Cloud, was efficiently completed within three weeks. Essentially, the work teams performed a direct mapping of the old provider's Kubernetes environment to Google Kubernetes Engine (GKE).

"The integration of Google Agent Development Kit (ADK) with Vertex AI was a game-changer for Myrmex's architecture. We managed to orchestrate multiple agents working in parallel and, by leveraging the cache optimization of these tools, we drastically reduced the need for data reprocessing."

Alexandre Fernandes — Head of Engineering, AINEXT

The backbone of AINEXT's infrastructure on Google Cloud is GKE in Autopilot mode. This choice was crucial as it abstracts the operational complexity of Kubernetes cluster management, automatically operating with scalability and security. The result was operational excellence, as the company could redirect 100% of its engineering efforts to Myrmex product innovation instead of spending time and labor on infrastructure management.

On the other hand, for Myrmex's development and intelligence, Google Cloud provides a cutting-edge ecosystem focused on AI. Vertex AI is the central platform, the AI core of Myrmex, used for the complete model lifecycle: training, deployment, inference, and RAG (Retrieval-Augmented Generation).

"The transition to Google Cloud was a leap in security maturity. We managed to orchestrate an architecture where Cloud Armor protects the perimeter, and end-to-end encryption protects the asset. But the differentiator lies in the intelligence layer: we apply advanced Guards directly at the inference stage, allowing us to deliver innovation with Myrmex without sacrificing rigorous control over data privacy and security," comments Vinicius Silva, Head of Cybersecurity at AINEXT.

The orchestration of AI models and the construction of Myrmex's agentic structure are accelerated by Google Agent Development Kit (ADK). The ADK is fundamental as it enables fluid integration with Vertex AI and simplifies the MLOps pipeline. The agent architecture is complex: when a user asks a question, Myrmex orchestrates multiple agents working in parallel to perform validations, analysis, and research, but returns the response to the user sequentially.

In addition to core tools, the infrastructure uses BigQuery for data analysis and monitoring, Cloud SQL for relational data, and Cloud Storage for storage.

Higher Platform Performance and More Control in Infrastructure Management

"The efficiency of integrated solutions led to cost reduction. But the biggest impact, certainly, was the acceleration of time-to-market and the improvement in product performance. Low latency and high availability of 99.5% ensure a robust and reliable experience for our clients."

Marcelo Benevides — CEO, AINEXT

The joint use of ADK with Vertex AI generated massive performance gains for AINEXT, and one of the biggest benefits was the drastic reduction in latency: the processing time for a final response in complex actions dropped from over 5 minutes to 1 minute and 20 seconds (~73% improvement). This advantage was obtained, in part, by cache optimization offered by both ADK and Vertex, allowing agents to store conversation context and avoid complete online data analysis at every interaction.

Conversely, for basic interactions, response time dropped from 2 minutes to 30 seconds, a 75% improvement. More expressive numbers correspond to the deployment speed of new AI agents, which increased by 45%, and agility in event response and data analysis, improved by up to 40%.

The use of AI solutions was a central benefit: the integrated ecosystem of Vertex AI and Agent Development Kit simplified the MLOps pipeline. Thus, the model lifecycle was boosted by over 35%. These gains and advanced cloud tools reinforce the real-time autonomous security value proposition of the Myrmex platform.

"With guaranteed infrastructure, AINEXT's strategy becomes even more supported by technological transformation. The company envisions expanding the capabilities of Myrmex agents, knowing it has the computational power of Google Cloud," comments Rafael Freire, CTO of the company.

For the future, AINEXT is also evaluating the adoption of AlloyDB due to its high-performance compatibility with PostgreSQL, in addition to using Cloud Spanner. Spanner, offering relational and vectorized data management, is seen as a database already prepared for the AI world - and fundamental for cybersecurity threat investigation.

AINEXT is the technology company responsible for developing Myrmex, a multi-agent AI platform for cybersecurity used to audit, operate, and document devices, infrastructure, applications, and users.

To learn more, read the full article on Google Cloud.