Security Researcher

Fortinet•Burnaby, BC

9h•Onsite

About The Position

Join Fortinet, a cybersecurity pioneer with over two decades of excellence, as we continue to shape the future of cybersecurity and redefine the intersection of networking and security. At Fortinet, our mission is to safeguard people, devices, and data everywhere. We are currently seeking a dynamic Security Researcher/Pentester to contribute to the success of our rapidly growing business. We are looking for a highly motivated individual who can thrive in a fast pace environment and successfully contribute to the team. You would act as a Security Researcher/Penetration Tester. This is a highly technical role, combining cutting-edge AI and LLM-powered attack techniques with traditional penetration testing skills. You will assist the Information Security leadership on evaluating the security posture of Fortinet Cloud Services. This position will be strongly focusing on AI powered red teaming, cloud security adversarial research and LLM exploitation.

Requirements

3+ years of dedicated experience in information security role with strong pen-tester practiced background (CVE, Pentest reports, Tech articles may be asked as proofs).
Professional penetration test skills on information technologies including operation systems, software frameworks, database, web applications and networks.
Strong knowledge of the fundamentals of web applications including. authentication, authorization, session management, HTTP protocol, web language, web server and browser architecture and implementation principle.
Proven skills with traditional pen-test tools (Nessus, Burp Suite, Nuclei, SQLmap).
Demonstrated practices of using Kali Linux / Metasploit to craft POCs for known vulnerabilities; ability to extend and automate these toolchains with LLM co-pilots or autonomous AI agents.
Hands-on experience with LLM fine-tuning frameworks and techniques: LoRA / QLoRA (Unsloth, LLaMA-Factory), supervised fine-tuning on security datasets, and direct preference optimization (DPO) for behavior shaping.
Proficiency in using LLMs and AI agents for offensive cloud security: automated IAM privilege-escalation discovery, cloud misconfiguration enumeration, AI-generated exploit payloads, and natural-language-driven attack orchestration.
Familiarity with AI/ML cloud service attack surfaces: prompt injection against retrieval-augmented generation (RAG) pipelines, embedding space attacks, model API abuse, and inference-time adversarial inputs.

Nice To Haves

Certifications such as OSCP , OSWE , HTB CPTS, HTB CWEE are highly valued.

Responsibilities

Conduct regularly penetration tests and security evaluations on Fortinet cloud products, covering web applications and server backend under various authentication level to identify vulnerabilities and security risks.
Investigating and Writing POCs for published vulnerabilities and help production team to evaluating the exploitability and risks.
Red team activities, specifically on developing AI-assisted scripts, agents, and programs to penetrate and infiltrate in-scope systems and ICT technologies; this includes leveraging LLMs as autonomous attack agents capable of chaining exploits across cloud environments.
Collecting threat intelligence, evaluating and maintaining traditional Pentest tools, and tracking emerging AI/LLM-based offensive security techniques and tooling.
Conduct AI red team exercises targeting LLM-integrated cloud services and AI APIs, including prompt injection, indirect prompt injection, jailbreaking, and model inversion attacks.
Perform LLM fine-tuning and abliteration research — including training uncensored or capability-unlocked variants of open-source models (LLaMA, Gemma4) via LoRA/QLoRA and representation-engineering techniques.
Develop and operate LLM-powered penetration testing pipelines: using AI agents (ReAct, tool-use) to automate reconnaissance, vulnerability enumeration, exploit generation, and post-exploitation chaining against cloud-native targets (Kubernetes clusters, serverless functions, IAM privilege escalation paths, cloud storage misconfigurations).
Research and document adversarial attack surfaces unique to AI-powered products: RAG pipeline data poisoning, embedding inversion, model supply-chain compromise, and training data extraction.