Skip to content

// platform & MLOps engineer

Anatoly (Till) Lazarev

A platform and MLOps engineer in Novi Sad, Serbia. I build the platforms other engineers ship on: ML infrastructure, Kubernetes, GPU, and the security underneath. At a fintech serving millions of users I built the company's ML platform from scratch and still run it as its sole engineer. This is my personal site — not a company.

A few numbers

7+
Years building platforms
5
Team workspaces on the ML platform I built
72
Concurrent workloads on 8 shared H100s
50%+
Cloud spend cut, company-wide

From systems I've built and run — zero to production, and operated end to end.

What I do

The platform, and what sits under it.

ML Platforms & MLOps

From data to a trained, versioned and served model in one place: notebooks, pipelines, experiment tracking, model registry and serving, built for several teams at once.

GPU Infrastructure

Shared GPU capacity instead of one card per person: VRAM slices, quotas that explain a refusal before launch, and nodes provisioned on demand per workload.

Kubernetes & Cloud

EKS and bare-metal, OpenTofu/Terraform, GitOps with Argo CD, multi-tenancy and policy. Reproducible infrastructure, defined in code.

Platform Security

SSO, Vault-issued credentials and internal PKI, admission policy, mTLS, and tenant isolation that holds up under review.

Developer & AI Tooling

SDKs, CLIs and MCP servers that let people and AI assistants use a platform from a laptop, without cluster access or cloud keys.

Observability & Reliability

Prometheus, VictoriaMetrics and Grafana at high cardinality, OpenTelemetry tracing, and alerts worth waking up for.

// contact

Anatoly (Till) Lazarev

Platform & MLOps engineer, based in Novi Sad, Serbia.

[email protected]

linkedin.com/in/lazarevtill

github.com/lazarevtill