Radu – Python, Kubernetes, Docker

A Move-based blockchain network that lets developers build fast, scalable decentralized applications. As the first SRE I built the company's entire AWS infrastructure from scratch and carried the network from devnet through testnet to two mainnet launches, while development moved fast and requirements kept changing.

• Bootstrapped the whole AWS setup (accounts, networking, IAM, compute, databases) with Terraform/CDKTF, following best practices from day one.
• Started deliberately simple, with EC2 provisioned by Terraform and configured with Ansible, which carried the network from devnet to the first mainnet. Once requirements settled, led the migration to Kubernetes on EKS.
• Wrote a custom Helm chart that ran our own nodes and let the community run their own nodes on AWS with the same configuration.
• Built CI/CD with GitHub Actions and introduced Nix with a Cachix binary cache for the Rust builds, cutting build times by about 50%.
• Deployed the Grafana LGTM observability stack (Loki, Grafana, Tempo, Mimir) with StatusCake and PagerDuty alerting, and managed Aurora Serverless PostgreSQL.
• Secured signing keys with HCP Vault (Transit) and supported the network through both mainnet launches.

Web3 account-abstraction infrastructure (ERC-4337 bundler and paymaster) that lets dApps sponsor gas and relay smart-account transactions on 40+ EVM chains. I led the move of the backend to cloud-native microservices on Kubernetes and modernised CI/CD and observability around it.

• Led the refactor of the backend into cloud-native microservices running on Kubernetes, managed with Terraform and Helm. I designed and prototyped the changes in Bash, and a developer colleague implemented them in TypeScript.
• Ran the bundler platform on GKE: per-chain autoscaled StatefulSets behind a load balancer, with RabbitMQ, Redis and MongoDB Atlas, serving 40+ EVM chains.
• Migrated CI/CD from Jenkins to GitHub Actions, with OIDC to GCP and Helm deployments.
• Set up monitoring and logging: Datadog APM, distributed tracing and SLOs with burn-rate alerting, plus Grafana Loki and Promtail for logs.
• Using Datadog's continuous profiler, found a memory leak in the TypeScript services and traced it to a logging library. The developers replaced it and the leak disappeared.
• Managed MongoDB Atlas, company networking (VPC, VPN, firewall), security best practices and cloud cost across AWS and GCP.

Core software for Polkadot and the Substrate blockchain framework. Developers had no easy way to test changes against a realistic multi-node network, so I led Simnet: a platform that spins up a full Polkadot/Substrate network on Kubernetes on demand, runs behaviour-driven test suites against it and tears it down. It later became Zombienet, which is still in use today.

• Designed the architecture around the Kubernetes operator pattern: a CRD describes the network and an in-cluster operator creates and reconciles the nodes.
• Built the developer experience the engineers asked for: a CLI and a simple config file instead of raw Kubernetes objects. I prototyped the CLI and all cloud interactions in Bash, then a developer colleague ported them to Rust and later TypeScript.
• Integrated it into GitLab CI so merge requests to main were tested against a real network.
• Ran a 300-node network for two weeks so developers could track down a memory leak in Polkadot, and applied chaos engineering to test reliability under failure.
• Built GitLab CI/CD pipelines with an efficient caching solution for Rust builds, and IaC with Terraform, Ansible, Helm and ArgoCD on GCP.
• The tool was adopted by many internal developers and external parachain teams.