Kayode – AWS, Kubernetes, Python

A self-hosted platform for securely executing untrusted, AI-generated code within private Kubernetes environments, enabling organizations to process sensitive workloads without relying on third-party sandboxing services. The project focused on building a scalable and secure execution engine with automated workload orchestration, lifecycle management, observability, and infrastructure hardening to support reliable code execution in production environments.

• Designed and built the control-plane API and reconciliation engine end-to-end, including state-machine modeling and crash-safe lease handling;
• Hardened the Kubernetes execution sandbox against common container-escape vectors;
• Wrote a 26-test suite (SQLite + Postgres) wired into CI, plus Prometheus instrumentation;
• Designed a Helm chart and GitOps-style deployment workflow (ArgoCD app-of-apps, PR-based promotion).

A multi-tenant Kubernetes platform for provisioning and managing isolated ephemeral environments for engineering teams. The project focused on scalable environment orchestration, automated lifecycle management, tenant isolation, request routing, and infrastructure automation, enabling efficient use of shared cluster resources while improving deployment reliability and operational scalability.

• Architected and built a multi-tenant Kubernetes platform from the ground up, defining isolation boundaries, autoscaling strategies, and ingress routing;
• Designed and implemented a TTL-based reconciliation service to automate environment lifecycle management and prevent resource sprawl across 50+ ephemeral environments;
• Reduced deployment failures by 90% by introducing automated conflict detection, retry mechanisms, and exponential backoff logic;
• Standardized deployment workflows by creating a reusable Helm-based service template that became the team's default deployment approach;
• Owned production operations for the platform, including on-call support, incident response, and the creation of operational runbooks.

A standardized Kubernetes deployment framework to eliminate inconsistencies across engineering teams and establish a common deployment approach. The project focused on creating reusable deployment templates with built-in best practices for resource management, health checks, labeling, and security, while integrating seamlessly with existing CI/CD pipelines. This enabled teams to deploy new services using standardized configurations instead of building deployment infrastructure from scratch, improving consistency, reliability, and operational efficiency.

• Designed and implemented a reusable parent Helm chart that embedded shared platform standards and best practices by default;
• Defined clear boundaries between standardized platform configurations (security settings, resource guardrails) and service-specific parameters to balance consistency with team autonomy;
• Restructured the CI/CD deployment pipeline to ensure new services automatically inherited standardized and secure deployment defaults;
• Drove the adoption of the deployment framework across engineering teams, establishing it as the default onboarding path for new services.

Contributed to the modernization of a highly regulated cloud platform by leading critical infrastructure and database transformation initiatives. The project focused on zero-downtime database migration, cloud infrastructure automation, CI/CD optimization, cost reduction, and disaster recovery planning, ensuring high availability, operational resilience, and compliance with strict data governance requirements.

• Led the zero-downtime RDS→Aurora migration using AWS DMS, with Terraform-provisioned infrastructure;
• Re-engineered the CI/CD pipeline (Jenkins), cutting deploy time from 4h → 1.5h;
• Reduced monthly infrastructure spend by 40%;
• Designed and validated a DR strategy meeting a 4-hour RTO / 1-hour RPO against strict resilience requirements;
• Authored Terraform modules for reproducible, auditable infrastructure provisioning.