Member of Technical Staff - Post-Training (Reinforcement Learning & Generalization)

Description

The company builds security-first foundation AI models and end-to-end products that help enterprises solve business problems. Its team includes researchers, engineers and designers.

In this Member of Technical Staff role in Post-Training, you will apply reinforcement learning to large language models and build methods that help them generalize beyond their training distribution. You will design the training recipes, environments and evaluations that show whether a model has learned a skill or only fitted its training tasks, then turn working approaches into the recipes behind production models. You will collaborate with the post-training data infrastructure, evaluation and pre-training teams.

Responsibilities

• Run large-scale RL post-training experiments, such as RLHF and reinforcement learning from verifiable rewards, and analyze the results.
• Develop methods for out-of-distribution generalization, including reward design, variety in tasks and environments, curricula and regularization.
• Create held-out evaluations for unseen tasks, domains and formats that expose reward hacking and benchmark overfitting.
• Find the causes of training problems such as instability, reward over-optimization and capability regressions.
• Convert research that works into reliable, reproducible training recipes for production models.

Requirements

• You bring strong research and software engineering skills and hands-on experience training LLMs.
• You have applied reinforcement learning to LLMs, such as PPO, GRPO, RLHF or RLVR, ideally at scale.
• You have worked on generalization, robustness or distribution shift, shown through publications or shipped models.
• You run rigorous experiments with clean ablations and careful controls, and you draw sound conclusions from noisy results.
• Experience with distributed training, such as JAX, PyTorch, FSDP or Megatron, is a plus.
• You communicate clearly and well.
• You understand data quality deeply. You enjoy working through datasets in detail, judge their quality by instinct, and can anticipate how different data characteristics will affect training outcomes.

Conditions

The role can be remote or based in one of the company's offices in Toronto (headquarters), London, New York City, San Francisco, Montreal, Paris, Berlin or Seoul. The company is remote-friendly and has no minimum in-office requirement.

Check the location listed on the posting for the core time zone or working hours expected.

More offices are opening soon.

The company welcomes applicants from all backgrounds and offers equal opportunities.

AI-enabled tools may be used to screen applicants against the role criteria. They help recruiters find candidates but do not limit what recruiters review.

Benefits

• Perks below apply to full-time employees.
• A weekly lunch stipend of $75/£75, or the equivalent in your local currency.
• Full health and dental benefits, plus a separate mental health budget.
• RRSP matching, a 401K and a pension scheme.
• 100% parental leave top-up for up to 6 months, for either parent.
• Annual enrichment benefits covering arts and culture, fitness and wellness, quality time, and a workspace improvement credit.
• An education and learning stipend for conferences, courses and coaching.
• 6 weeks of paid vacation, which is 30 working days.
• A travel budget to visit other offices if you work remotely, plus an annual company offsite.
• For office-based staff, a daily lunch program, snacks and regular community and social events.
• For those not near an office, a co-working benefit to work alongside others in their city.
• A $500 home office stipend to set up your workspace.