Opportunity at
SynthesiaStaff Research Engineer - Multimodal Generative Modelling
Synthesia, an artificial intelligence video platform provider established in 2017 and backed by over $530 million in total funding, seeks a Staff Research Engineer to focus…
Career Tools
About This Role
Synthesia, an artificial intelligence video platform provider established in 2017 and backed by over $530 million in total funding, seeks a Staff Research Engineer to focus on multimodal generative modelling. Operating out of London with global teams, the company builds enterprise visual communication tools utilized by more than ninety percent of Fortune 100 businesses. In this position, you will join the Voice team within an R&D department of over forty…
Job Description
Synthesia, an artificial intelligence video platform provider established in 2017 and backed by over $530 million in total funding, seeks a Staff Research Engineer to focus on multimodal generative modelling. Operating out of London with global teams, the company builds enterprise visual communication tools utilized by more than ninety percent of Fortune 100 businesses. In this position, you will join the Voice team within an R&D department of over forty professionals, contributing to the long-term goal of developing human-interactive models that perceive and respond to user actions and emotions.
The role centers on creating voice-to-voice models capable of simultaneous text and speech production, low-latency conversational streaming, and natural interaction without traditional turn-taking delays. You will partner with colleagues to shape the research roadmap, propose novel architectures, conduct end-to-end training, and transition models into production environments. Successful candidates will possess strong software engineering capabilities, a solid background in generative modelling and transformers, and practical experience taking deep learning models from inception to live deployment.
Responsibilities
- Formulate research roadmaps to introduce new model capabilities for customers across short and long timeframes
- Propose novel multi-modal system architectures focusing on text and voice
- Develop and evaluate streaming conversational systems for low-latency interactive voice and video synthesis
- Design solutions that enhance emotional expressiveness and natural human-computer interaction
- Implement designs across all stages from pretraining through post-training
- Integrate and test neural codecs, diffusion models, and flow-matching architectures to improve realism
- Establish latency-aware and interaction-based evaluation metrics for conversational systems
- Monitor ongoing research developments in audio-visual diffusion, autoregressive models, and multimodal large language models
- Curate new datasets to supplement existing data sources
- Manage post-training initiatives including direct preference optimization, fine-tuning, and distillation
- Deploy models into production with optimized runtimes and incorporate user feedback
Requirements
- Ability to introduce novel ideas and designs advancing interactive multimodal systems
- Strong understanding of generative modelling applied to sequential or multimodal data
- Hands-on experience with large language models or transformer-based architectures
- High proficiency in PyTorch including distributed training and model optimization
- Solid grasp of time-series modeling and tokenization in audio, speech, or video
- Demonstrated capability to prototype rapidly, test hypotheses, and iterate efficiently
- Proven experience executing end-to-end deep learning model training from data preparation to evaluation
- Strong general software engineering skills for contributing to shared research infrastructure
Qualifications
- Experience shipping generative models into live products used at a meaningful scale
- Background working on conversational or interactive systems where latency and responsiveness were primary constraints
- Experience with large-scale training of large language models leading to reasoning capabilities
- Experience owning research problems end-to-end from architecture proposal to production deployment
- History of cross-modal or cross-team collaboration to deliver unified systems
- Familiarity with real-time or streaming architectures
- Familiarity with audio and speech generation architectures like diffusion models, neural codecs, flow-matching, or autoregressive decoders
- Excellence in voice, text, or video modalities
- Evidence of original research contributions through publications or open-source work at top-tier venues such as NeurIPS, CVPR, ICML, ICLR, or Interspeech
Core Skills
Frequently Asked Questions
Answers are based only on the employer’s listing; where it doesn’t say, neither do we.
What is the employment type for this role?
The posting specifies that this is a full-time position.
Where is the job located?
The specific location for this role is not stated in the job posting, though Synthesia is headquartered in London with offices across Europe and the US.
What is the salary for this position?
The job posting does not specify salary details.
What department will I join?
You will join the Voice team within Synthesia's R&D department, which consists of over forty people.
Sample Interview Questions
AI-generated questions tailored to this specific role — a preview of the full practice set.