Founding Machine Learning Engineer

Numi Technologies Ltd.; Dubai; On-site; AED 300,000 - 600,000; Full time

Job summary

Numi Technologies Ltd. is seeking an experienced ML engineer to own the migration from deterministic logic to a learned system end to end in Dubai. You will implement transformer-based representation learning over raw transaction sequences and drive production-grade ML pipelines.

The role requires 5+ years of production experience in self-supervised/representation learning and a strong background in tabular ML (LightGBM, XGBoost). In-person, five days a week in the Dubai office.

Qualifications

• 5+ years of production experience in self-supervised learning, representation learning, or transformer-based sequence modelling.
• You have trained models, not just called them.
• Self-supervised and representation learning expertise.
• Strong tabular ML as a baseline discipline (LightGBM, XGBoost).
• Advanced degree in a quantitative discipline or equivalent commercial work.

Responsibilities

• Own the migration from deterministic logic to a learned system end to end.
• Develop transformer-based representation learning over raw transaction sequences.

Skills

Self-supervised learning

Transformer models

Representation learning

Tabular ML (LightGBM/XGBoost)

5+ years production ML experience

Education

Advanced degree in a quantitative discipline

Tools

LightGBM

XGBoost

Job description

We're an in-person team. If you're in Dubai, you're in the office 5 days a week. For the right person, we'll make an exception: set up wherever, and we'll fly you in a few times a year and cover the costs.

About the role

Numi's Decision Engine works. Today, it's a very well-tuned pile of maths. You'd own the migration from deterministic logic to a learned system end to end. Transformer-based representation learning over raw transaction sequences.

What you'll need

• 5+ years of production experience in self-supervised learning, representation learning, or transformer-based sequence modelling
• You've trained models, not just called them
• Self-supervised and representation learning expertise
• Strong tabular ML as a baseline discipline (LightGBM, XGBoost)
• Advanced degree in a quantitative discipline or equivalent commercial work