Staff Software Development Engineer (LLM)

Fortinet•Sunnyvale, CA

4d•$196,500 - $219,300

About The Position

We are seeking a Staff Software Development Engineer specializing in Large Language Models (LLMs) to join our team. In this role, you will be instrumental in architecting and implementing robust security measures for LLM applications. You will build scalable systems to monitor and filter LLM traffic in real-time, protect against prompt injection attacks, and prevent unauthorized data leakage. Your work will involve developing advanced monitoring, logging, and alerting systems to detect anomalies and potential misuse. You will collaborate with cross-functional teams to translate security requirements into platform features, mentor junior engineers, and ensure the timely delivery of high-quality software. Effective communication with both technical and non-technical stakeholders is crucial.

Requirements

Hands-on experience with deploying or integrating large language models or other AI/ML systems (e.g. implementing model inference pipelines, fine-tuning models, or working with LLM APIs and prompt handling). Strong understanding of how prompts and context are managed in LLM applications.
Solid knowledge of application security principles and experience building secure systems. Familiarity with AI-specific threats (prompt injection, data poisoning, output manipulation) and a keen interest in staying ahead of new generative AI attack vectors.
Proven experience designing microservices architectures and using containerization (Docker) and orchestration (Kubernetes). Comfortable with cloud platforms (AWS, GCP, or Azure) and designing observable, resilient services in a production environment.
Ability to design clean, efficient, and secure APIs. Strong understanding of network protocols, data caching, and performance optimization.
Knowledge of data protection and privacy best practices – able to design systems that handle sensitive data responsibly and comply with regulations. Understanding of responsible AI principles (ethics, bias, transparency) and how they relate to secure AI system design.
Familiarity with emerging AI security guidelines such as OWASP’s Top 10 for LLMs/Generative AI Security (e.g. knowledge of prompt injection, insecure output handling, data poisoning risks) and experience implementing related mitigations.
Experience with retrieval-augmented generation (RAG) architectures, vector databases/embedding stores, or streaming data pipelines for ML – especially as they relate to monitoring and securing LLM workflows (helps in addressing vector embedding attack vulnerabilities).
Understanding of responsible AI and techniques for detecting AI-generated misinformation or hallucinations. Experience building or integrating content filtering, policy enforcement, or fact-checking systems in AI applications is a plus.
Strong programming and debugging skills, particularly in Python and C/C++.
Familiarity with Frameworks like PyTorch or TensorFlow for model integration; libraries such as Hugging Face Transformers or LangChain for LLM and prompt management; LLM APIs (OpenAI, etc.) and vector databases is beneficial.

Responsibilities

Architect and implement functions to monitor and filter LLM requests/responses in real time, preventing prompt injection attacks and unauthorized data leakage.
Build a highly scalable pipeline capable of handling high-volume LLM traffic with low latency, including optimizing databases and caching for quick threat detection and response.
Develop monitoring, logging, and alerting systems to detect anomalies in LLM usage (e.g. suspicious spikes indicating denial-of-service attacks or unusual prompt patterns indicating misuse).
Collaborate with teams to translate security requirements into platform features. Mentor junior engineers on secure backend development and best practices in an Agile environment.
Ensure the timely delivery of high-quality software features while adhering to project schedules.
Communicate effectively across teams, with both technical and non-technical stakeholders, in both verbal and written forms.