Research Scientist
About the role
Role focused on building and leading post-training and RL methods for agents using production traces to create environments and rewards, with experiments at scale and production deployment.
Responsibilities
- Post-train language models on tasks and environments derived from production data.
- Advance methods across RL, distillation, and learning from user feedback.
- Design evaluations to test generalization beyond training distribution and treat evals as research problems.
- Build synthetic environments and data pipelines that convert messy production traces into training signals.
- Take large-scale experiments end-to-end from hypothesis through analysis to a shipped model.
- Set research direction: decide what’s worth training, what’s worth measuring, and what to discard.
Qualifications
- Serious ML research experience, ideally with post-training, RL, or alignment.
- A visible research track record: publications, open implementations, or usable trained models.
- Rigorous experimental design ensuring experiments answer their intended questions.
- Fluency in Python and a modern ML stack (PyTorch or JAX), including distributed training.
- Comfort working with messy real-world data.
- Ability to go from paper to prototype to production impact independently.
- Desire for research to run in production.
Skills
About ThirdLayer
Building Dex, the AI Browser Copilot
ThirdLayer is an applied research and product lab building the next interface for human–AI collaboration. Our first product is Dex, an AI copilot for the browser. Think Cursor for knowledge work, optimized for speed and embedded directly where you work. Most web agents are built for set-and-forget tasks. But real work is fluid and messy—spread across tabs, tools, and evolving decisions. Dex transforms your browser into an intelligent, multimodal workspace that understands your intent through voice, text, and action. It enhances familiar interfaces, automates tedious steps, and removes the friction of handoffs. Beneath the surface, Dex is more than a browser agent — it’s a foundation for: ● Long-term, cross-modal memory grounded in real workflows ● A new interaction layer to replace static UI with a collaborative canvas ● Personalization: understanding how individuals approach tasks and navigate software We're building the infrastructure for adaptive, intent-aware systems that work the way you do.
More open roles at ThirdLayer
- 2mo ago
- Staff/Senior Engineer, AI Platform (Dex)2mo agoRemote · Engineering
- 2mo ago
- Customer Enablement Engineer4mo agoHybrid (Global) · Engineering
- Product Engineer4mo agoEngineering