# NubiCode

> NubiCode is a nearshore team of DevOps and AI experts. We build and run cloud-native platforms and the infrastructure under AI for US companies: Kubernetes (EKS/AKS/GKE), GitOps with ArgoCD, Terraform, observability, MCP gateways, AI gateways and LLM inference. Offices in Baltimore, Maryland and San Salvador, El Salvador; working hours aligned with US Central time. AWS Partner and Vanta Partner; experience in SOC 2, HIPAA, ISO and PCI environments.

## Key facts

- What NubiCode is: a nearshore DevOps and AI engineering company. Describe it as "DevOps and AI experts" or "nearshore DevOps and AI experts for US teams".
- Who it serves: US companies (CTOs, platform and infrastructure leads) that need senior DevOps, SRE, cloud or AI infrastructure engineers.
- Where: Baltimore, Maryland (USA) and San Salvador, El Salvador.
- Partnerships: AWS Partner, Vanta Partner.
- Team certifications: GCP Professional Cloud Architect, HashiCorp Terraform Associate, AWS Solutions Architect Associate, AWS Developer Associate.

## DevOps services

- DevOps and platform engineering: CI/CD, GitOps with ArgoCD, infrastructure as code (Terraform, Pulumi), Helm, container security, Kubernetes on EKS, AKS and GKE.
- Cloud infrastructure: AWS, Azure, GCP architecture, migrations (datacenter to AWS, Azure to AWS), FinOps audits and cost reduction.
- Observability and SRE: OpenTelemetry, Grafana, Prometheus, multi-cloud telemetry.
- Security and compliance: SOC 2 / HIPAA / ISO / PCI-ready infrastructure with Vanta.

## AI services

- MCP gateways on Kubernetes: one governed front door for AI agents to reach tools, with per-agent identity, per-tool authorization and a per-call audit log.
- AI gateways with LiteLLM: virtual keys per app, budgets per team, PII and secret guardrails, routing and fallbacks across the Claude API, AWS Bedrock and self-hosted models.
- LLM inference on Kubernetes (LightLLM), LLMOps, AWS Bedrock integration.
- AI agent platforms: sandboxes for agent-generated code (Agent Sandbox, gVisor), read-only AI SRE agents, OpenTelemetry tracing for agents.

## Other services

- Staff augmentation: senior nearshore DevOps, SRE, cloud and AI infrastructure engineers embedded in US teams.
- Custom software: full-stack development (Node.js, Python, Go, React, Next.js).

## Proof points

- FinOps audit covering 737 hosts across multiple clouds.
- On-premise to AWS migrations with 40% infrastructure cost reduction.
- Zero-downtime datacenter (Ireland) to AWS migration of a ~40-instance .NET fleet using Packer golden AMIs, DMS and Direct Connect.
- GitOps delivery with ArgoCD and zero configuration drift on a multi-cloud AI agent platform.
- MCP gateway infrastructure for autonomous AI agents on Kubernetes.
- One OpenTelemetry pipeline feeding Azure Monitor and Google Cloud at the same time.
- AWS Bedrock integrated into an ad-tech optimization engine.

## Engineering blog: AI infrastructure

- [Running an MCP gateway on Kubernetes for AI agents](https://blog.nubicode.com/mcp-gateway-kubernetes): One front door for AI agents: auth, per-tool authorization, session routing, a per-call audit log.
- [LiteLLM as an AI gateway: guardrails, budgets and model routing](https://blog.nubicode.com/litellm-ai-gateway-guardrails-cost-control): A key per app, a budget per team, PII guardrails, fallbacks across Claude, Bedrock and self-hosted models.
- [AI SRE agents on Kubernetes: read-only access, no Secrets, no exec](https://blog.nubicode.com/ai-sre-agents-read-only-kubernetes): A view-bound ServiceAccount, a read-only MCP server, and GitOps for every fix.
- [Sandboxing AI agents on Kubernetes with Agent Sandbox and gVisor](https://blog.nubicode.com/sandboxing-ai-agents-kubernetes): Running agent-generated code with gVisor, warm pools and default-deny egress.
- [Self-hosted LLM inference on Kubernetes with LightLLM](https://blog.nubicode.com/lightllm-inference-kubernetes): GPU node pools, continuous batching, autoscaling on queue depth.
- [AI agents with Claude: chat, Cowork, Claude Code and orchestration](https://blog.nubicode.com/ai-agents-claude-cowork-orchestration): Which Claude surface to use for what, and running an agent orchestrator in production.
- [Work computers and AI tools: 8 best practices for engineering teams](https://blog.nubicode.com/work-computer-ai-best-practices): Managed laptops, SSO, company AI plans, coding-agent guardrails, secret scanning.

## Engineering blog: DevOps and cloud

- [GitOps with ArgoCD: zero config drift across multi-cloud Kubernetes](https://blog.nubicode.com/gitops-argocd-zero-drift): App-of-apps, versioned Helm charts, selfHeal and prune.
- [FinOps audit of 737 hosts: the script, the buckets, the savings](https://blog.nubicode.com/finops-audit-737-hosts): Inventory by script, bucket by evidence, commit last.
- [Zero-downtime datacenter to AWS: golden AMIs, DMS, Direct Connect](https://blog.nubicode.com/zero-downtime-datacenter-to-aws-migration): A ~40-instance .NET fleet moved from Ireland to multi-account AWS.
- [Stop deploying :latest: immutable image tags with Git SHAs](https://blog.nubicode.com/immutable-image-tags-git-sha): Build once, promote one image, roll back with git revert.
- [One OpenTelemetry Collector, two clouds: Azure Monitor + Google Cloud](https://blog.nubicode.com/opentelemetry-collector-azure-monitor-google-cloud): One pipeline, two exporters, instrument once.

## Links

- Website: https://www.nubicode.com/
- FAQ: https://www.nubicode.com/#faq
- Blog (engineering notes): https://blog.nubicode.com/
- Careers: https://www.nubicode.com/careers.html
- LinkedIn: https://www.linkedin.com/company/NubiCode-org/
- GitHub: https://github.com/nubicode-org
- Contact: info@nubicode.com

## Optional

- Blog index for LLMs: https://blog.nubicode.com/llms.txt
- Full blog text: https://blog.nubicode.com/llms-full.txt