Senior AI Engineer
תיאור המשרה
We're looking for a Senior AI Engineer to own those environments and the evaluation that decides which tasks are good enough to train a frontier model.
What you'll do RL environments Build RL environments end to end: the replica systems, the planted attack, and the verifier that grades the run Own environment and verifier design for tool-use, web, and computer-use agents Make environments airtight: sandboxed, free of eval-awareness cues, and closed to reward hacking Read rollout traces, find where the agent took a shortcut, and close it Task evaluation Own Alice's proprietary task evaluator, the engine that decides whether a task is ready for a frontier lab Make it the industry's sharpest quality bar for RL tasks: realism, threat-model accuracy, and training value Keep it learning: every customer review and every post-training run makes it stricter Run infrastructure Scale the run pipeline: queue-driven runners on Kubernetes, per-run sandboxes, and a grader the agent can't reach Package environments so they drop into customer training and evaluation pipelines Working with Research Work with researchers as one team, from a task idea to an environment that's ready to train on Turn what post-training runs reveal into fixes in the environments Review designs and code, and raise the engineering bar for the people around youRequirements: What we're looking for 6+ years as a backend engineer, building production systems BSc in Computer Science or a related field, preferably from a leading university Strong Python, and production experience with Docker, Kubernetes, and cloud infrastructure (AWS preferred) Experience designing distributed systems and message queues A track record of owning systems from design to production High autonomy in an undefined space. You take a loose problem and come back with working code Fluent written and spoken English Advantage Academic background in AI or Machine LearninggreenTxtBg!: a degree, thesis, or research Hands-on AI engineering: building with LLMs, agents, or evaluations Exposure to post-training: RL fine-tuning, reward design, or rollout analysis * Background in security, red teaming, or trust and safetyThis position is open to all candidates.
מתעניינים במשרה הזו?