Applied Scientist — Adversarial ML & Detection

About the Client

Our client is a global enterprise software provider serving more than 1,500 customers across financial services, government, media, and IT.

The company provides enterprise-grade PostgreSQL solutions and managed cloud services. Its solutions help organizations modernize applications, migrate databases from legacy systems, and operate across hybrid and multi-cloud environments.

The platform unifies transactional, analytical, and AI workloads while ensuring security, compliance, and high availability, with built-in AI capabilities.

About the Team

You will join the Governance team, responsible for building detection models for adversarial and unreliable agent behavior — injection, jailbreak, poisoned-input, and hallucination detection.

Responsibilities

• Train and harden classifiers on adversarial data for injection/jailbreak and poisoned-input detection.

• Build hallucination detection and reasoning-hardening techniques.

• Contribute detection signal into audit log confidence scoring and the escalation queue.

Requirements

• Adversarial ML experience (attack detection, robustness, classifier training on adversarial data).

• Strong expertise in PostgreSQL and solid SQL skills.

• Security mindset and comfort working in fully audited environments.

• Fluent English, both written and spoken.

Nice to Have

• NLP / embedding-space depth — semantic similarity attacks, embedding-based anomaly detection

Location

• Fully remote.

• Candidates must be based in the EU or the Americas.

What We Offer

• Fully remote work.

• A professional, supportive, and friendly team.

• Long-term employment with competitive compensation based on experience.

• Continuous knowledge sharing with engaged co-workers.