Autonomous configuration optimization platform powered by Reinforcement Learning. It systematically explores millions of configuration permutations — Kubernetes resource limits, JVM parameters, HPA thresholds — to hit strict SLO targets while minimizing infrastructure costs. Operates both offline in pre-production and live in production environments.
Akamas is a configuration optimization platform that uses reinforcement learning to systematically explore large configuration search spaces and converge on settings that satisfy defined performance objectives at minimum resource cost. Given a goal (latency target, throughput, or cost) and a search space (JVM heap flags, Kubernetes resource limits, HPA thresholds, GC parameters), Akamas runs controlled experiments across thousands of permutations.
Two operating modes: pre-production load test mode drives automated experiments in staging; live shadow mode observes production traffic while running parallel optimization without a staging environment. The output is a concrete configuration diff validated against the target SLO — not a list of suggestions.
Enterprise-only pricing; all tiers require sales engagement. Integrates with Dynatrace, Datadog, and New Relic for baseline collection and performance validation.
Key Features
Reinforcement learning optimization: systematically explores millions of permutations across JVM flags, Kubernetes resource limits, HPA thresholds, and GC parameters to converge on configurations meeting defined SLO objectives
Pre-production load test mode: drives automated load tests across configuration variants in staging, validating results before any production change is applied
Live shadow mode: observes production traffic patterns while running optimization experiments in parallel, enabling optimization on realistic workloads without a staging environment
Objective-driven output: produces a concrete configuration diff validated against the target SLO — not a recommendation list requiring manual evaluation and risk assessment
Multi-technology search spaces: optimizes JVM-based services (heap, GC), Kubernetes resource requests and limits, HPA scaling thresholds, and database connection pool settings in a single study
APM integration for validation: connects to Dynatrace, Datadog, and New Relic to collect performance baselines and validate optimization results against defined SLO thresholds
Integrations
5 total
orchestration
Kubernetes
monitoring
Datadog
cloud
AWSGCPAzure
Pricing
4 tiers
Kubernetes & Microservices
—
$2,900/credit/month; up to 5 microservices concurrently, live production + offline optimization
