---
title: "Senior Adversarial ML & Agent Runtime Security Engineer"
slug: "senior-adversarial-ml-agent-runtime-security"
id: "dab7f703-4047-402b-8c3b-60c2be025603"
department: "Machine Learning"
department_code: "ML"
location: "Zurich, Switzerland"
workplace_type: "Hybrid"
employment_type: "Full-time"
compensation: "CHF 145k – CHF 190k (annual package)"
is_remote: false
is_featured: false
headcount: 1
canonical_url: "https://jobs.talent247.co/jobs/senior-adversarial-ml-agent-runtime-security"
markdown_url: "https://jobs.talent247.co/jobs/senior-adversarial-ml-agent-runtime-security.md"
apply_endpoint: "https://jobs.talent247.co/api/public/jobs/senior-adversarial-ml-agent-runtime-security/apply"
schema_endpoint: "https://jobs.talent247.co/api/public/jobs/senior-adversarial-ml-agent-runtime-security/form-schema"
webmcp_tool: "apply_for_role"
updated_at: "2026-11-15T17:23:58.221Z"
content_signal: "search=yes, ai-input=yes, ai-train=no"
---

# Senior Adversarial ML & Agent Runtime Security Engineer

## Position Telemetry

- **Reference:** `SENIOR-ADVERSARIAL`
- **Practice / Sector:** Machine Learning (`ML`)
- **Location:** Zurich, Switzerland
- **Workplace Type:** Hybrid
- **Employment Type:** Full-time
- **Compensation:** CHF 145k – CHF 190k (annual package)
- **Canonical HTML URL:** https://jobs.talent247.co/jobs/senior-adversarial-ml-agent-runtime-security

## Role Dossier & Specification

THE CLIENT & MANDATE
Our client is an acclaimed AI security pioneer headquartered in Zurich, spun out of Switzerland's world-renowned machine learning and cybersecurity ecosystem. As enterprises transition from passive chatbots to autonomous agents granted write access across internal databases, codebases, and APIs, our client provides the real-time runtime defense layer and continuous red-teaming engine keeping those agents safe against indirect prompt injection, jailbreaks, and tool hijacking.

They have retained Talent247 to appoint a Senior Adversarial ML & Agent Runtime Security Engineer at their Zurich headquarters. Sitting on their Threat Research & Guardrail Modeling team, you will co-develop both offensive automated red-teaming agents that uncover novel foundation-model vulnerabilities and sub-15ms defensive classifier architectures deployed inline across global enterprise traffic.

KEY RESPONSIBILITIES
• Research, train, and distil ultra-fast multilingual transformer and encoder/decoder guardrail models detecting indirect prompt injection, toxic content, PII leakage, and unauthorized tool-call trajectories under strict <15ms p99 latency SLAs.
• Build evolutionary and RL-driven offensive red-teaming agents that automatically probe newly released closed- and open-weight LLMs and multi-agent harnesses for novel jailbreaks and cross-context exploits.
• Curate massive multilingual adversarial datasets leveraging red-team telemetry from hundreds of thousands of attack simulations alongside verified enterprise edge cases.
• Minimize false-positive friction on benign developer, code, and security-analyst prompts through hard-negative synthesis, threshold calibration, and domain-adaptive fine-tuning.
• Collaborate closely with Rust runtime engineers to quantize (INT8/FP8) and compile compact classifier heads for both CPU sidecars and GPU clusters.
• Publish threat intelligence advisories, open benchmarks, and peer-reviewed AI security research representing our client across the European and global safety community.

TECHNICAL ENVIRONMENT & OFFERING
Their stack includes Python, PyTorch, ONNX/TensorRT distillation, vLLM red-team fleets, Rust inference gateways, and multi-GPU training clusters. Working hybrid in central Zurich, our client offers CHF 145,000–CHF 190,000 base salary plus generous equity, Swiss pension and benefits, and full Swiss work-permit/relocation sponsorship.

## Core Requirements

ESSENTIAL QUALIFICATIONS
• 5+ years of experience in machine learning engineering, NLP, or adversarial AI security research with production model deployment ownership.
• Deep proficiency in PyTorch, encoder/decoder transformer architectures, knowledge distillation, quantization-aware training, and synthetic adversarial data generation.
• Practical understanding of LLM and agent security threat vectors: direct/indirect prompt injection, RAG poisoning, MCP/tool-use hijacking, and multi-turn jailbreaks.
• Strong experimental rigor around extreme precision-at-high-recall tradeoffs, ROC calibration, and out-of-distribution robustness across dozens of languages.
• Fluent English communication and collaborative engineering habits across ML research and systems teams.

PREFERRED / BONUS EXPERIENCE
• PhD or MSc in Machine Learning, Cybersecurity, or NLP, or publications/CTF track record in adversarial ML and LLM red-teaming.
• Experience optimizing sub-20ms transformer inference on CPU (AVX-512/AMX, ONNX Runtime) or GPU (TensorRT/Triton).

## Application Schema & Direct Apply (`apply_for_role`)

Candidates and authorized AI agents on behalf of a candidate can submit an application via the declarative WebMCP form tool `apply_for_role` on https://jobs.talent247.co/jobs/senior-adversarial-ml-agent-runtime-security or via HTTP `POST https://jobs.talent247.co/api/public/jobs/senior-adversarial-ml-agent-runtime-security/apply` (`application/json` or `multipart/form-data`). Non-browser HTTP agents must include header `X-Talent247-Intake-Token` obtained from `GET https://jobs.talent247.co/api/public/intake-token` (or automatically returned in `GET https://jobs.talent247.co/api/public/jobs/senior-adversarial-ml-agent-runtime-security/form-schema`).

### Form Fields

- `full_name` (text, required): Full name
- `email` (email, required): Email
- `phone` (phone, optional): Phone
- `location` (text, optional): Location
- `linkedin_url` (url, optional): LinkedIn URL
- `cover_letter` (textarea, optional): Cover letter
- `consent` (checkbox, required): I consent to my data being processed for this application
- `cv_file` / `cv_base64` + `cv_filename`: Upload a CV in .pdf or .docx format (<= 10 MB) or provide an `https://` LinkedIn profile URL in `linkedin_url`.

### Alternative: Talent Pool Intake (`join_talent_pool`)

If this mandate is not an exact match, register once in the Talent247 Talent Pool at https://jobs.talent247.co/talent-pool (`https://jobs.talent247.co/talent-pool.md`) via WebMCP tool `join_talent_pool` or `POST https://jobs.talent247.co/api/public/talent-pool`.
