Senior Infrastructure/DevOps engineer at Menlo Research. I work on the systems that sit underneath our research and product teams — hybrid cloud spanning on-prem OpenStack and Kubernetes, GPU clusters, and the major hyperscalers, all glued together with GitOps.
Core contributor to Jan, an open-source local LLM desktop app. Mostly on release engineering, CI/CD, and the boring parts that make shipping reliable across macOS, Windows, and Linux.
Currently working on
- Multi-tenant Kubernetes on bare metal (Kamaji + Cluster API)
- GPU observability with DCGM, Prometheus, and Grafana
- Keycloak / OIDC identity across our internal platform
- Argo CD + Image Updater for everything that can be a Git commit
Tech I work with
- Cloud: AWS, GCP, Azure, and on-prem — wired together as one hybrid platform
- Orchestration: Kubernetes (on-prem via CAPI/Kamaji, GKE/EKS/AKS), OpenStack (Kolla-Ansible), Proxmox
- GitOps & CI/CD: Argo CD, Argo CD Image Updater, GitHub Actions, Helm, Kustomize, OCI registries
- Observability: Prometheus, Grafana, Loki, Tempo, Pyroscope, DCGM for GPU
- Identity & access: Keycloak, OIDC/SSO, Google Workspace integration, 1Password ESO
- Networking: Cisco Nexus vPC, pfSense HA, Headscale/Tailscale, Cilium, Cloudflare (Workers, Access, Tunnels)
- Storage & data: Ceph / Rook, TimescaleDB, PostgreSQL, Valkey/Redis, S3-compatible object stores
- Edge & realtime: LiveKit/WebRTC, STUN/TURN, Cloudflare Workers
- Languages: Go, Python, Bash, and a healthy amount of YAML
Open to chat about Hybrid cloud architecture, running GPU workloads on-prem vs. in the cloud, self-hosted observability and identity stacks, and homelab over-engineering.
📍 Hanoi, Vietnam · 📫 Reach me via Linkedin