About the team
The Dubbing and Localisation team owns Synthesia's end-to-end video translation product, transforming existing videos into new languages while preserving voice, lip sync, and timing. As the team scales, building robust, automated infrastructure for measuring and maintaining translation quality is important.
The role
You'll own the design and scaling of an automated translation quality evaluation framework — building the infrastructure that lets the team measure quality consistently, automatically, and at scale. You'll be partnering closely with the team to ensure the systems you build reflect real-world product quality needs.
What you'll work on
- Designing and building scalable evaluation infrastructure for automated translation quality measurement
- Owning the rollout of that infrastructure starting with Localisation, then scaling across Dubbing and other teams
- Driving consistency and automation in how quality is measured across the pipeline
- Partnering with the wider organisation to align evaluation infrastructure with real product and user quality signals
- Maintaining and iterating on evaluation systems as the pipeline and product evolve
- Working closely with Synthesia's internal R&D team on evaluation methodology and tooling
What we're looking for
- Strong ML infrastructure engineering background
- Hands-on experience designing and scaling ML evaluation or quality measurement systems in production
- Solid understanding of ML pipelines and how models are evaluated
- Comfortable owning systems from architecture all the way through to delivery
- Strong backend engineering fundamentals; this role ships real production systems, not research prototypes
- Experience working cross-functionally across multiple teams with different requirements
- Familiarity with NLP, translation quality metrics, or speech/audio evaluation is a strong plus
Nice to have
- Experience with LLMs applied to quality scoring or translation evaluation
- Background in async or batch processing pipelines
- Prior work in localisation, dubbing, or media processing tooling
The hiring process
- 30-40min call with our Technical Recruiter
- 45min call with an Engineering Manager about your past projects
- 90min Live Coding interview, a collaboration exercise with our Engineer (not Leetcode style)
- 60min System Design, technical discussion usually about a real problem we had in production
- 45min call with leadership, focus on examples of collaboration, growth and going through details of your future team