Guardrail ML Engineer for Prompt Security

Fortinet, Inc.; Sunnyvale (CA); On-site; USD 150,000 - 183,000; Full time

Benefits offered by this job

Medical insurance

Dental insurance

Vision insurance

Life and disability insurance

401(k)

11 paid holidays

Vacation time

Sick time

Equity program

Job summary

Fortinet, Inc. is hiring a Machine Learning Engineer to own detection models for FortiAIGate. You will train, evaluate, optimize, and serve models in production, building guardrails against prompt injection and sensitive data exposure. You’ll work with PyTorch, Transformers, and modern inference stacks.

The role emphasizes quantization, distillation, and deployment on Triton/vLLM with GPU optimization, plus threat research for evasion. Candidates must be authorized to work in the U.S.

Qualifications

• Strong Python and production PyTorch experience.
• Experience training, fine-tuning, and evaluating transformer models.
• Production experience with modern inference serving systems.
• Experience with quantization, distillation, pruning, or graph compilation.
• Ability to design test sets and reason about precision/recall.
• Familiarity with containerized deployment and MLOps practices.

Responsibilities

• Build and train guardrail models for prompt and data safety.
• Design a tiered detection cascade with latency targets.
• Work across encoder and decoder model families with distillation.
• Quantize, distill, and compile models for deployment on GPU appliances.
• Harden models against evasion and develop regression tests.
• Own evaluation suites and monitor model drift in production.

Skills

Python

PyTorch

Transformer models

Hugging Face

Go/Rust/C/C++

Inference serving

quantization

distillation

Docker

Kubernetes

CUDA

ONNX Runtime

TensorRT

vLLM

Triton

Tools

ONNX Runtime

TensorRT

Triton Inference Server

vLLM

Job description

Fortinet, Inc. is hiring a Machine Learning Engineer to own detection models for FortiAIGate. You will train, evaluate, optimize, and serve models in production, building guardrails against prompt injection and sensitive data exposure. You’ll work with PyTorch, Transformers, and modern inference stacks.

The role emphasizes quantization, distillation, and deployment on Triton/vLLM with GPU optimization, plus threat research for evasion. Candidates must be authorized to work in the U.S.