Machine Learning Lead

Cyber Valley GmbH; Tübingen; On-site; EUR 90,000 - 130,000; Full time

Benefits offered by this job

Competitive salary

Remote within Germany

Small, fast-moving team

Job summary

MATVIS, a leader in secure AI agents, is seeking a Machine Learning Lead to own end-to-end detection features and drive production-grade ML systems in a security-conscious environment.

You will tackle threat surfaces like prompt injection, PII leakage, and data exfiltration, ship containerized services in Docker and Kubernetes, and collaborate in a small, fast-moving team in Germany or remote within Germany.

Qualifications

• Full MLOps ownership from data to production.
• Strong Python and PyTorch experience.
• Security mindset with pentesting/CTF/adversarial ML.
• Benchmarking and evaluation discipline.
• Clear written communication and collaborative work style.
• Docker and REST APIs proficiency.

Responsibilities

• Own core detection and analysis features end to end: dataset collection, training, evaluation, deployment, and everything that breaks afterwards.
• Work across the full threat surface: prompt injection, jailbreaks, PII detection and redaction, content filtering, data exfiltration, agent misuse.
• Build the parts of the detection stack that are not models: traffic analysis for Policy Autopilot, parametrization, and the business logic that turns a signal into a decision.
• Build and defend the benchmark harness. Detection rate, false positive rate and latency are product metrics here.
• Ship containerized services into Docker and Kubernetes environments.
• Debug across the full stack, from model output through runtime, container and network.

Skills

Python

PyTorch

MLOps

Docker

REST APIs

English fluency

Security mindset

Benchmarking

Communication

Tools

Kubernetes

Job description

MATVIS is looking for a Machine Learning Lead to join their team!

About MATVIS

MATVIS secures AI agents. They inspect traffic inline, on any model and any infrastructure, and stop prompt injection, PII leakage, data exfiltration and shadow AI before they reach anything that matters. Model-agnostic, ISO 27001 certified, and deployed inside their customers' own environments so their data never leaves.

Founded in Tübingen in 2023. Enterprise customers run them in production today, most of them working against AI Act and NIS2 deadlines, so they can deploy AI with confidence.

Your role

As one of their two Machine Learning Leads, you will:

• Own core detection and analysis features end to end: dataset collection, training, evaluation, deployment, and everything that breaks afterwards.
• Work across the full threat surface: prompt injection, jailbreaks, PII detection and redaction, content filtering, data exfiltration, agent misuse.
• Build the parts of the detection stack that are not models: traffic analysis for Policy Autopilot, parametrization, and the business logic that turns a signal into a decision.
• Build and defend the benchmark harness. Detection rate, false positive rate and latency are product metrics here, and customers feel all three.
• Design security into the pipeline itself. They build their detection infrastructure in house, from data provenance through to their own training and evaluation tooling.
• Ship containerized services into Docker and Kubernetes environments.
• Debug across the full stack, from model output through runtime, container and network.

Must have

• Full MLOps ownership. You have taken a model from raw data into production and kept it running.
• Strong Python and PyTorch.
• A security instinct: pentesting, red teaming, CTF, bug bounty or adversarial ML. If you can attack systems and train models, they are interested regardless of what your degree says.
• Benchmarking and evaluation discipline.
• Clear written communication and a collaborative, adaptable way of working. Small team, high async load.
• Docker and REST APIs.
• Fluent English, written and spoken.

Nice to have

• Experience attacking LLM applications that were already in production.
• German.

What they offer

• Competitive salary based on industry standards.
• Tübingen, in the middle of Cyber Valley, next to the Max Planck Institute, ELLIS and the Tübingen AI Center. Or remote within Germany.
• A team small enough that what you build ships within weeks.

MATVIS is committed to diversity and inclusion. They strive to build a diverse team and welcome applications from people of different backgrounds, skin color, religion, sexual identity, gender, disability or age. Follow themon LinkedIn to stay up to date with what's happening at MATVIS.

We send out an overview of the most important news and events in the Cyber Valley Community. Only available in English.