Staff / Principal Research Scientist - UK
Inworld · Remote · UK ·
COMPENSATION
€•• – ••k
Listed on the employer's job board (Ashby)
- Staff / Principal level
- Fully remote (UK)
- Foundation models & RL
- No visa sponsorship
- £140k - £200k base
Join Inworld's research lab to build first-party speech models and serve LLMs for high-volume realtime workloads. You will tackle ambiguous problems in model research, inference infrastructure, and developer APIs, supporting millions of daily users across various industries.
What you'll do
- Conduct research on foundation models, new architectures, and scaling strategies
- Design and implement evaluation benchmarks and quality measurement loops
- Explore frontier topics such as multimodal models, agents, and world models
- Own the full research cycle from framing questions to shipping results
- Collaborate on large-scale serving infrastructure and modular APIs
What you'll need
- Strong background in ML/NLP with practical experience or a PhD
- Experience with training, RL, reward modeling, or scaling foundation models
- Ability to thrive in ambiguity and define your own research direction
- Proven track record of building, breaking, and understanding complex systems
- Legal right to work in the United Kingdom is mandatory
Nice to have
- Published research at top conferences like ICML, ICLR, or NeurIPS
- Public work including open-source contributions or non-trivial side projects
- Experience with LLM-as-judge, failure analysis, or test-time compute
- Interdisciplinary experiments or competition achievements outside academia
Stack
- LLMs
- Speech models
- Realtime inference
- Modular APIs
How you'll work
- Fully remote within the United Kingdom
- Full-time employment
- Flat structure with minimal process overhead
- Fast iterations and impact-focused culture
What you get
- Base salary between £140,000 and £200,000
- Equity compensation included
- Comprehensive benefits package
- Potential US visa and relocation support for SF Bay Area
Hiring process
- No cover letter required
- Submit a link to something you have built or researched
- Evaluation based on evidence of work rather than credentials