I'm a Principal SRE and Platform Engineer. I build cloud-native infrastructure and developer tooling, mostly in Go, for environments where reliability, security, and compliance actually matter.
- Tooling that removes toil: Go CLIs and automation that give teams self-service instead of tickets
- Observability: OpenTelemetry, Prometheus, distributed tracing, and SLIs/SLOs that tell you something true
- Infrastructure as code and GitOps: Terraform, event-driven provisioning, and Argo CD-style GitOps (Git as the source of truth, progressive delivery with Argo Rollouts) so deployments need no hands
- Security by default: service mesh, mTLS, hardened Kubernetes, and secrets handled properly
- Reliability: incident response, runbooks, and migrations done without drama
- Standardizing observability and platform patterns across large fleets of services
- Exploring MLOps and applied GenAI: RAG design, grounded answers, and treating LLM accuracy as a testable, regression-gated concern rather than a vibe
- Building guardrails for AI coding agents: keeping sensitive data out of prompts without breaking the workflow
ocloud: an open-source Go CLI for Oracle Cloud Infrastructure operations that I use daily.
cui-proxy: a fail-closed local proxy that masks sensitive data (IPs, resource IDs, credentials, your own dictionary) in AI coding-agent traffic with realistic synthetic values, then restores them in responses. It has a strict block-by-default policy profile, a live dashboard, and a macOS menu-bar app. Install with brew install rozdolsky33/tap/cui-proxy.
OpenLex: a proof-of-concept retrieval-augmented legal research assistant (NY landlord-tenant law), built to learn MLOps by running an LLM system like any production service. Hybrid retrieval (pgvector + Postgres FTS), citations that answer only from retrieved statute, and a CI-gated accuracy harness: 21 golden questions with known-correct citations, run on Haiku in CI and Sonnet on demand, with a live evaluation report showing per-question pass/fail and model divergence. SRE regression-gating and cost-awareness applied to GenAI.
Also: SRE mentor, and a former musician who now finds rhythm in reliable systems. πΌ