Job Description

Principal / Director, AI Research – Reinforcement Learning for LLMs

We're hiring a Principal or Director-level AI Researcher with deep expertise in Reinforcement Learning and LLM post-training to join our growing AI research group. This is a research-first role, with a mandate to push the frontier of model alignment, safety, and performance — working with foundation models in real-world, high-stakes environments.

You won’t be handed toy problems or legacy systems. Instead, you'll lead applied research efforts focused on tuning, aligning, and optimizing large models for privacy, security, and interpretability - in one of the few spaces where LLMs have both massive scale and measurable consequences.

What You’ll Work On:

This role centers on building and refining intelligent agents that interact with sensitive data and complex access controls, using modern reinforcement learning and post-training techniques:

• Post-training of LLMs using RL: Design and run experiments with methods like PPO, DPO, RLAIF, and other fine-tuning strategies to align model behavior with security and privacy goals
• RL for Self-Correction & Redaction: Enable models to iteratively improve their predictions on document classification, redaction, and identity resolution through self-rewarded feedback loops
• Model Alignment & Safety: Contribute to the development of our “LLM Firewall” — filtering prompts/responses to prevent jailbreaking, data leakage, and adversarial exploits
• Inference Stack & Optimization: Collaborate with engineers optimizing our in-house inference stack to make LLaMA-class models performant at scale

What We’re Looking For:

• Demonstrated expertise in Reinforcement Learning applied to language models or decision-making agents
• Strong understanding of post-training methodologies (e.g., RLHF, DPO, preference modeling, rejection sampling, offline RL)
• Solid background in LLMs , token-level reasoning , and language modeling internals
• Publication record or research contributions in top-tier venues (NeurIPS, ICLR, ICML, ACL, etc.) preferred
• Ability to work independently and iterate quickly — experience in scrappy, high-output research environments a plus
• Industry experience is not required — we care more about the depth of your research thinking and experimentation rigor

Why This Role:

• Join a company with massive real-world data , impactful use cases, and a mature infrastructure
• Avoid the grind of infra-focused roles — we’ve already solved those problems
• Shape the next phase of LLM alignment , self-correcting models , and AI safety at inference time
• Work on problems with technical depth and direct product impact

Job Tags

Similar Jobs