DeepL4 дня назад
Research Manager | Production Inference
Зарплата не указана
Полная занятостьLondon
Обязанности
- 01Lead and develop a high-performing team of research scientists and ML engineers, building strong development plans, fostering a candid and non-retaliatory feedback culture, and maintaining high standards of technical rigour and delivery
- 02Own the team's research and development roadmap for production inference systems, in close collaboration with senior ICs and cross-functional stakeholders, balancing near-term reliability commitments with longer-horizon research bets on inference efficiency and architecture
- 03Act as the primary technical interface between the Production Inference team and adjacent functions including foundational models research, voice research, applied research, infrastructure, and product ensuring research output is well-scoped, well-communicated, and delivered without creating downstream bottlenecks
- 04Drive the reliability, efficiency, and cost performance of DeepL's model serving stack, including strategic decisions around serving infrastructure evolution (load balancing, autoscaling, runtime selection, and hardware utilisation)
- 05Operate with a high degree of autonomy, defining the team's direction and pushing for results in an environment where requirements from product or commercial stakeholders can be ambiguous or evolving
- 06Play an active role in identifying, assessing, and recruiting research and engineering talent as the team continues to develop
Требования
- 01PhD (preferable) in Computer Science, Mathematics, Physics, or a comparable quantitative discipline, or possess a strong ML/systems background with equivalent research depth
- 02Strong foundation in production ML systems, inference optimisation, or model serving at scale
- 03Proven experience leading a team of researchers or ML engineers, with a track record of developing talent, maintaining delivery rigour, and holding the balance between research quality and production reliability
- 04Comfortable operating across the full model lifecycle from training handoff through to production deployment, monitoring, and efficiency improvement
- 05Understand infrastructure and compute constraints without needing to own them directly
- 06Excellent communication skills and the ability to translate complex technical direction into clear goals for both technical and non-technical stakeholders
- 07Solution-oriented and decisive, able to define direction and drive outcomes