Ingénieur(e) IA en déploiement avancé, Travel Intelligence// Forward Deployed AI Engineer, Travel Intelligence
Montréal - Remotefull time
Remote
$360k–$425k
H1B: yes
Hopper Forward Deployed AI Engineer at Hopper responsible for ensuring AI safety, security, and readiness in partner deployments. Owns guardrails, adversarial testing, launch readiness, and partner security reviews.
Responsibilities
Hopper powers travel for some of the world's largest banks, airlines and travel companies, and our AI agents act on their customers' behalf by looking up accounts, changing and cancelling bookings, and handling payments and loyalty points. As the Forward Deployed Engineer, AI Trust and Reliability, you make sure that AI is safe, secure and ready for production in every partner deployment. You work in a small forward-deployed team as the person who owns guardrails, adversarial testing, launch readiness and partner security reviews, and you are the bridge between our partners' security, risk and IT teams and Hopper's engineering and security teams.
What would your day-to-day look like
Tailor guardrails to each partner's policies, regulations, risk appetite and languages, covering what the AI may do, what it must refuse and when it hands off to a human
Design protections for sensitive data and actions, including data masking, identity checks and customer consent before any change to a booking or payment
Tune guardrails so they stop real risks without blocking legitimate customers
Red-team every deployment before launch and after significant changes, testing for jailbreaks, prompt injection, data leakage, unauthorized actions and social engineering
Build and maintain automated adversarial test suites that run continuously, in every language a partner serves
Turn findings into fixes with engineering, and into regression tests so they stay fixed
Lead partner security questionnaires, architecture reviews and AI or model-risk reviews, and respond to partner penetration-test findings
Document how our AI handles data, access and decisions in terms a regulator or auditor would accept
Own the go-live checklist for each deployment, covering functional, regression, safety and load testing, sign-off criteria and rollback plans
Extend our automated testing across products, so every model, prompt or configuration change is tested before it reaches customers
Qualification
Required
Experience running full post-training loop end to end
Experience with supervised fine-tuning and at least one of preference optimization or reinforcement learning
Strong Python skills and command of training stack such as PyTorch, Hugging Face, and vLLM
Experience running jobs on rented or customer GPUs
Experience building benchmarks and graders from real-world data