About
Joe Sindel
Senior Site Reliability Engineer. Most recently Major Incident Commander @ Coinbase. Building self-hosted AI on the side. Open to new roles.
12+ years building, scaling, and protecting production systems. Most recently I operated one of the world’s largest proof-of-stake validator fleets at Coinbase and served as Major Incident Commander for company-wide SEV events — coordinating ops, comms, and engineering across war rooms to drive MTTR down and protect customer-facing trading availability.
What I Do
- Incident Command — Led live SEV-1/SEV-2 response (IMAG model) across trading, custody, and crypto products. Ran blameless post-incident reviews and owned follow-through so the same class of incident doesn’t recur — making reliability work a budgeted, scheduled outcome rather than a recurring fire drill.
- Site Reliability Engineering — Embedded with product teams to set SLOs/SLIs, raise the production-readiness bar on new launches, and improve observability so on-call engineers detect issues before customers do.
- Blockchain Infrastructure — Operated PoS validator infrastructure at scale, including Ethereum — where every minute of downtime is lost yield and slashing is a financial event. Tuned bare-metal hosts, redesigned beacon node topology, and carried the on-call pager as protocol SME.
- Platform Migrations — ECS→EKS, GitOps with Flux/ArgoCD, IaC standardization on Terraform/Packer, CI/CD modernization (GitHub Actions, Jenkins), and Reserved Instance strategies that have delivered ~50% AWS savings.
- Security & Compliance — SOC 1 / SOC 2, HIPAA, security council facilitation, end-to-end encryption — translating audit findings into prioritized engineering work instead of letting them rot in a tracker.
What I’m Building Outside of Work
- Thor — Private, self-hosted multi-backend AI inference platform on NVIDIA Jetson AGX Thor (Blackwell sm_110, 128 GB unified memory). Three concurrent inference backends — vLLM, Ollama, and TensorRT-LLM — served behind a unified OpenAI-compatible API, with voice chat over cellular via Tailscale + Caddy and LLaVA-13B vision-language object detection feeding an in-progress autonomous drone perimeter-sentinel.
- GoNFTme — Zero-fee Web3 crowdfunding platform with NFT rewards on Base (L2). End-to-end design and build: Solidity contracts on OpenZeppelin/Hardhat, Next.js 15 + TypeScript frontend, Wagmi wallet integration, and a security-first posture (Zod, SonarQube, OWASP review, unit + E2E tests) before mainnet.
Recent Writing
- What Major Incident Command Actually Buys You — what IMAG buys you at company scale, where it gets misapplied, and how to make the post-mortem loop actually close
- One-to-Many Beacon Nodes: Rethinking PoS Validator Topology — why the rigid 1:1:1 default breaks at scale, and the bare-metal tuning underneath that determines whether your validators attest on time
- Building a Private AI Server with NVIDIA Jetson AGX Thor — three inference backends, voice chat over cellular, custom observability, all on-prem
Stack I Reach For
AWS (EKS, ECS/Fargate, VPC, IAM, S3, CloudFront), Kubernetes, Terraform, Packer, Helm, GitOps (Flux, ArgoCD), Datadog/Prometheus/Grafana/Netdata, Python, Bash, TypeScript, Go (working), Solidity, systemd, Tailscale/WireGuard, Caddy. Active hands-on work in vLLM, Ollama, TensorRT-LLM, AWQ/INT4 quantization, and CUDA/Blackwell on NVIDIA Jetson.
Let’s Connect
Always interested in discussing reliability, incident response, blockchain infrastructure, or self-hosted AI. Especially open right now to senior SRE / infrastructure roles.
- 📧 Email: joesindel@gmail.com
- 💼 LinkedIn: linkedin.com/in/joesindel
- 💻 GitHub: github.com/josephsindel
- 📬 More on the contact page
“The best way to learn is by doing, and the best way to remember is by teaching.”