Engineering Manager – Production Inference

About the role

DeepL is looking for an Engineering Manager to lead the Production Inference team, responsible for delivering reliable, low‑latency model serving at scale. You will guide a group of research scientists and ML engineers, shaping both technical direction and people development.

Key responsibilities

• Lead and develop a high‑performing team, creating development plans and fostering a feedback‑rich culture.
• Own the research and development roadmap for production inference systems, balancing reliability commitments with long‑term efficiency research.
• Act as the primary technical liaison between Production Inference and adjacent functions such as foundational models, voice research, infrastructure, and product.
• Drive reliability, efficiency, and cost performance of the model serving stack, influencing load‑balancing, autoscaling, runtime selection and hardware utilisation.
• Identify, assess, and recruit research and engineering talent as the team grows.

Required profile

• Proven experience leading researchers or ML engineers, with a track record of talent development and delivery rigour.
• Strong background in Computer Science, Mathematics, Physics or a comparable quantitative discipline, or equivalent ML/systems expertise.
• Excellent communication skills, able to translate complex technical direction for both technical and non‑technical stakeholders.
• Solution‑oriented and decisive, capable of defining direction without waiting for higher‑level guidance.

Required skills

• Production ML systems
• Inference optimisation
• Model serving at scale
• LLM inference
• Speculative decoding
• Quantisation
• Serving infrastructure (load balancing, autoscaling, runtime selection)
• GPU inference runtimes and hardware utilisation

What we offer

• Diverse, internationally distributed team across more than 90 nationalities.
• Hybrid work schedule with flexible hours and two days per week in the office.
• Virtual Shares giving employees a stake in DeepL’s growth.
• Regular in‑person team events and monthly full‑day hack sessions.
• 30 days of annual leave plus mental‑health resources.