Description

I build high‑performance AI inference infrastructure.

I help companies deploy, optimize, and monitor LLMs (Llama, Qwen, vLLM) on GPU clusters with low latency and high throughput.

What makes me different:

• I write Rust, not Python, for inference pipelines — lower overhead, more predictable performance.

• I built `inferscope`, a Rust profiler that correlates client‑side token latency with per‑GPU metrics (NVML, SM utilization).

• I wrote a Kubernetes operator in Rust that treats cold start as a readiness signal — services become Ready only when the GPU is warm.

• I investigated NVIDIA Dynamo's KV‑aware router and quantified the load‑shedding trade‑off under saturation.

Typical deliverables:

✔ Inference platform on Kubernetes (vLLM/Triton, GPU autoscaling, GitOps with ArgoCD/Helm)

✔ Performance diagnosis (TTFT/TPOT, GPU occupancy, kernel tracing with eBPF)

✔ Infrastructure as Code (Terraform, Helm, ArgoCD)

✔ Custom observability stacks (Prometheus, Grafana, OpenTelemetry)

Industrial background (9 years in operations) gives me capacity planning, cost discipline, and cross‑functional communication. I work fully remote, European time zone (CET).

Let's talk about your inference bottlenecks.

Industry field of expertise

Languages

Italian
Native or bilingual
English
Conversational
French
Basic

Workplace preferences

Remote only

Primarily works remotely

specialty contract manufacturer
Industrial Operations Analyst
January 2018 - Today (8 years and 6 months)
Italy
Quantitative analysis for a mid-market specialty contract manufacturing operation: cost-by-workcenter modelling, order-level P&L, production-line productivity monitoring, capacity planning, and new-SKU pricing. Primary bridge between production, commercial, and finance leadership — and between the business and IT/development. This domain depth is what later grounded the optimisation and systems work.
AI Automation Strategic planning Kubernetes
Insurance Broker
Risk & Insurance Consultant (SMB segments)
January 2017 - January 2018 (1 year)
Northern Italy Pipeline System, Castel San Pietro Terme, BO, Italy
Technical risk assessment and policy structuring for industrial SMBs. First exposure to structured risk modelling, which later informed CVaR-based quantitative risk analysis in operations.