Performance Engineer: Containerized ML & Serverless GPUs

Verda; Helsinki; Hybrid; EUR 90,000 - 130,000; Full time

Benefits offered by this job

Healthcare

Lunch

Wellbeing

Equity

Job summary

Verda in Helsinki, Finland is hiring a Performance Engineer to own the surface for Verda's container and serverless GPU platforms. You will profile time-to-first-token, training step time, and latency, turning insights into concrete platform improvements across storage, compute, and networking.

You’ll work with cross‑functional teams, benchmark real workloads, and publish internal write-ups so engineers and customers understand trade-offs. Hybrid work in Helsinki with on-site presence.

Qualifications

• Production experience optimizing ML/AI workloads.
• Strong Linux storage and container runtime internals.
• Experience with distributed or network filesystems and with S3-compatible object storage.
• Familiarity with model-serving runtimes and data formats (e.g., safetensors, GGUF).
• End-to-end systems thinking across NIC, switch, filesystem, cache, container runtime, and GPU.

Responsibilities

• Profile and optimize the end-to-end path for containerized ML/AI workloads.
• Design and tune the storage layer between object stores and GPU nodes.
• Drive wins in time-to-first-token, training step time, and latency.
• Benchmark real workloads and translate findings into platform changes.
• Collaborate across compute, networking and platform teams.
• Publish internal write-ups to share trade-offs with the org and customers.
• Stay current with the evolving ML/AI ecosystem.

Skills

Performance optimization for ML/AI

Linux storage internals

Container runtime internals

Distributed/network filesystem

S3-compatible storage experience

Model-serving runtimes familiarity

End-to-end systems thinking

Go/Rust/Python experience

Tools

Go

Rust

Python

S3 tooling

Job description

Verda in Helsinki, Finland is hiring a Performance Engineer to own the surface for Verda's container and serverless GPU platforms. You will profile time-to-first-token, training step time, and latency, turning insights into concrete platform improvements across storage, compute, and networking.

You’ll work with cross‑functional teams, benchmark real workloads, and publish internal write-ups so engineers and customers understand trade-offs. Hybrid work in Helsinki with on-site presence.