Uses

What I run

The tools and stacks I reach for daily — mix of production infrastructure work at Careem and the AI tooling I use to ship faster.

AI assistants & agentic tooling

  • Claude Code — primary daily driver for engineering work. Multi-agent orchestration, custom skills, MCP-connected workflows. Daily user.
  • Cursor — in-IDE pair-programming when I want to stay in the editor surface.
  • GitHub Copilot — inline completion for boilerplate-heavy code.
  • Gemini — second opinion on prompts and longer-context reasoning tasks.
  • Model Context Protocol (MCP) — building production MCP servers for the AI Agent Platform at Careem. Wrote the primer on retrofitting microservices.

Cloud & Kubernetes

  • AWS EKS — production daily. Managed control plane, custom data plane (Karpenter, ARM/Spot fleets).
  • Karpenter — consolidation, Spot interruption handling, custom NodePools and NodeClasses. The 40% EKS cost cut leaned heavily on this.
  • VPA — rightsizing in steady-state workloads after Karpenter handled the bin-packing.
  • ArgoCD — GitOps for everything. At Intech I wired it to Keycloak for auditable Helm-based deploys.
  • Helm + Kustomize — both, picked per workload shape.
  • operator-sdk (Go) — built a bespoke Kubernetes Ingress Operator for bare-metal clusters at Intech.

Networking & gateway

  • Kong — consolidated 12 legacy gateways onto one Kong cluster at ~10k RPS with 99.99% uptime during cutover. Built a GitOps-driven bespoke control plane for route management.
  • Linkerd — service mesh for mTLS and observability without the ops cost of Istio.
  • Flagger — progressive delivery / automated canaries on top of Linkerd.
  • OpenResty (LuaJIT) — dynamic proxying + JWT bearer auth back at Intech; still a tool I reach for when nginx alone won't cut it.

Observability

  • Prometheus + Thanos — long-retention, high-cardinality metric storage across the EKS fleet.
  • Grafana — SLO dashboards, cost-tracking, on-call runbooks.
  • Google SRE golden signals (latency / traffic / errors / saturation) as the default mental model for instrumentation.

Infrastructure as code & CI/CD

  • Terraform + Terragrunt — multi-environment, multi-account provisioning. Took new service provisioning from days to minutes at Careem.
  • Ansible — bare-metal cluster provisioning and config drift remediation.
  • Docker — every workload, every environment.
  • GitLab CI / GitHub Actions — both, depending on the org. This site deploys via GH Actions with --ignore-scripts on npm ci as a supply-chain mitigation.

Languages

  • Go — operators, tooling, anything that needs to run as a tight binary.
  • Python — the glue. Agent platform tooling, automation scripts.
  • Java — legacy service surfaces and Spring-based identity providers from the Intech era.
  • C# / .NET — from the Alachisoft days building NCache and NosDB internals. Still reach for it when working in mature .NET shops.

This site

  • Astro with the minimal template, static output, zero JS shipped by default except for the theme toggle.
  • Tailwind CSS v4 via the Vite plugin, with custom @theme tokens.
  • Inter Variable self-hosted as a single woff2.
  • GitHub Pages via GH Actions, deployed in ~36 seconds end-to-end.
  • OG card generated programmatically with Playwright from scripts/og-template.html. Source: github.com/codemug/codemug.github.io.

Inspired by the long-running uses.tech community.