Senior Research Scientist (LLM post training)

PolyAI automates customer service through lifelike voice assistants that let customers lead a conversation. Our voice assistants make it possible for businesses to deliver outstanding customer service that rivals their human agents. Our customers, which include the world’s leading logos, are expanding how they use our platform, driving automation of critical customer service operations and integrating PolyAI into their daily customer service workflows.

We are looking for a Senior Research Scientist to join our world-class team and lead cutting-edge work on large language model (LLM) post-training. This is not just about applying standard fine-tuning techniques - it's about building the future of dialogue systems with novel approaches to reasoning, reinforcement learning, audio-first LLMs, and more.

As a Senior Research Scientist at PolyAI, you’ll lead impactful research projects from ideation through to deployment. You’ll be driving innovation in how we train and adapt LLMs for real-world conversations - spanning voice, text, and multimodal contexts. You'll work on frontier techniques such as:

• Conversational reinforcement learning
• Streaming and continuous turn-taking
• Audio-native LLMs
• Distillation of reasoning models
• Long-context
• Efficient adaptation

You’ll also play a key role in shaping the scientific direction of our research, mentoring junior colleagues, and collaborating cross-functionally to bring research into production.

Responsibilities:

• Lead and execute complex research projects with clear business impact.
• Design and implement novel post-training strategies including preference tuning, reward modeling, and synthetic supervision.
• Develop innovative model architectures and training approaches for conversational AI, including speech-aware and multimodal models.
• Conduct empirical studies to assess model performance in live deployments and iterate quickly based on real-world data.
• Generate, collect, and annotate training data - including synthetic and real-world conversational datasets - with an eye for quality and bias mitigation.
• Design robust evaluation metrics and benchmarks for LLM-based assistants in customer service domains.
• Work closely with engineering and product teams to integrate research into production environments.
• Collaborate with legal and compliance teams to ensure responsible use of data and models.
• Stay current with academic and industry advances in LLMs, ASR, TTS, RLHF, and multimodal learning.

Requirements:

• PhD in Machine Learning, Natural Language Processing, Computer Science, or a related field.
• 5+ years of hands-on experience in deep learning.
• Proven track record of research innovation, including published work or deployed systems.
• Strong programming skills in Python and deep learning frameworks like PyTorch.
• Demonstrated expertise in at least one domain area such as reinforcement learning, conversational AI, audio modelling, or LLM alignment.
• Experience leading projects end-to-end, from ideation to deployment.
• Excellent communication skills with the ability to write clear technical documents and explain complex concepts to diverse audiences.
• Comfortable working in ambiguity and driving clarity through experimentation and data.

Preferred Qualifications:

• Experience with speech technologies such as ASR and TTS.
• Familiarity with cloud environments (AWS, GCP, Azure).
• Exposure to RLHF, reward modelling, or human preference data collection.
• Prior work on real-time systems, streaming inference, or memory-efficient model deployment.

PolyAI Benefits - UK

💰Participation in the company’s employee share options plan

🏝25 days holiday, plus bank holidays

🏡Flexible working from home policy plus a one-off WFH allowance when you join

🌎Work from outside of the UK for up to 6 months each year

🧡Enhanced parental leave

📚Yearly learning budget

🚲Bike2Work scheme

🏥Private healthcare and dental cover, discounts on gym members and relaxation apps, and access to a range of mental health programs

Interview Process

1. Quick intro call with Talent (15 - 30 mins max)

2. Interview with Hiring Manager, 45 mins

3. Loop stage: 1x coding and 1x ML, both 45 mins

4. Final with CTO