Staff AI Research Scientist

About the role

As a Staff AI Research Scientist, you'll set the technical direction for the models behind GabForge — the assistant and the family of plugin agents that millions of everyday users will rely on. You'll own research bets from hypothesis to production, not just papers.

This is a senior, hands-on role. You'll spend your time training and evaluating models, mentoring engineers, and deciding where we should and shouldn't spend compute.

What you'll do

• Lead post-training (fine-tuning, preference optimisation, distillation) of our self-served models.
• Design evaluation suites that actually predict real-world quality, not just benchmark scores.
• Push the cost/quality frontier so our free tier stays free — smaller models, smarter routing.
• Partner with the inference team to ship research into a product running on hardware we own.
• Set the research roadmap and mentor a growing team.

Why it matters

We run our own metal and promise a permanent free tier. That makes efficiency a research problem, not an afterthought — your work directly decides what we can give away.

What we're looking for

• 8+ years in ML/AI research with a track record of shipped models, not only publications.
• Deep expertise in LLM training: pre-training or post-training, RLHF/DPO, distillation, quantisation.
• Strong PyTorch; comfortable owning a training run end to end on limited hardware.
• A bias for efficiency — getting more from a 12B model rather than reaching for a 70B.
• Experience mentoring senior engineers and setting technical direction.

Nice to have

• Publications at NeurIPS / ICML / ACL.
• Experience with on-device or single-GPU inference constraints.
• Open-source contributions.