Deep Learning Quantitative Researcher

About the role

Deep Learning Quantitative Researcher based in Dubai, responsible for designing and building core deep learning pipelines for applied quantitative alpha research, from data preparation to production deployment.

Key responsibilities

• Design and build the firm’s core deep learning pipelines for applied quantitative alpha research—from data preparation and distributed training through evaluation and production deployment.
• Drive the research agenda using applied deep learning techniques, owning the full empirical loop: problem formulation, model design, training, validation, and performance attribution.
• Uphold rigorous research discipline in a low signal‑to‑noise domain—strict out‑of‑sample hygiene, leakage prevention, and honest benchmarking against simpler baselines.
• Act as the firm’s central point of deep learning expertise: advise on architecture selection and training diagnostics, review model designs, and set standards for model evaluation and promotion.
• Facilitate seamless flow of model fitting and computation across teams through standardized training and inference interfaces and reusable components.

Required profile

• Top‑tier academic background from a globally top‑20 university (e.g., MIT, Harvard, Princeton, Stanford, Caltech) with PhD‑level training in Computer Science, Engineering, Physics, Mathematics, or Statistics.
• 3–5 years of professional experience applying deep learning to large‑scale problems, ideally in quantitative finance, with proven end‑to‑end ownership of a deep learning model lifecycle.
• Strong foundations in statistics, optimization, and machine learning theory; gold medal in a national or international olympiad is strongly preferred.

Required skills

• Python
• Modern deep learning frameworks (e.g., TensorFlow, PyTorch)
• Distributed/multi‑GPU training and mixed‑precision techniques
• Experiment‑management tools (tracking, hyperparameter optimization)
• C++ or CUDA optimization (optional)
• Large‑scale dataset handling (columnar formats, streaming loaders, point‑in‑time‑correct construction)