Machine Learning Engineer

Not everything is a chat box. Our clients need demand forecasts for freight, risk scores for lending, ranking for a marketplace, and pricing that holds up when a vendor changes their feed at 6am. That work needs a model with numbers behind it.

You'd take a problem from a client conversation to something running in production: the data, the features, the model, the monitoring, and the moment six months later when the world shifts and the model stops being right.

Agents handle the boilerplate and turn notebooks into services in an afternoon, which leaves you the part that needs a statistician.

• Frame the problem. Turn a business question into something measurable, and say when the answer is a heuristic instead of a model.
• Build the thing. Features, training, validation, and a baseline you have to beat before anything clever is allowed.
• Ship it into a product. An endpoint or a job that fits the engineering team's world, not a notebook with a README.
• Watch it age. Drift, retraining, and alerting on the quiet degradation nobody notices until a client does.
• Explain it. To the client's analysts, in terms they can challenge and check.

• Four Years of Applied ML: with models in production that someone depended on commercially.
• scikit-learn and Gradient Boosting: XGBoost or LightGBM, and the judgment to start there before reaching for anything deeper.
• PyTorch When It Earns It: sequence models and embeddings for the problems a tree can't hold.
• Experiment Tracking: MLflow, Weights and Biases or equivalent, so a result can be reproduced next quarter.
• Pulls Your Own Data: SQL you write yourself, and a query plan you can read when the extract takes an hour.
• Validation That Holds Up: cross-validation, leakage hunting, and backtests with time respected.
• Serving in Production: batch scoring or an API, containerized, versioned, and reversible.
• Drift and Retraining: inputs and outputs monitored, with a trigger for when the model needs newer data.
• Python Fluency: plus the engineering habits to make your work maintainable by someone else.
• Statistics You Can Defend: validation, leakage, class imbalance, and why your offline metric flattered you.
• Feature Pipeline Realist: you have dealt with training and serving skew and know how it sneaks in.
• Monitoring Mindset: you plan for the model getting worse from the day it ships.
• Fluent English, Written and Spoken: most of our work happens in a written thread or on a call, and both have to be clear. Arabic is a genuine advantage across the region.
• AI in the Loop: agents accelerate your code, and you stay the one accountable for the math.

• Skeptical of your own results: the first good number is a bug until proven otherwise.
• Business-minded: you care whether the model changed a decision, not whether it beat a benchmark.
• Pragmatic: logistic regression that ships beats a transformer that doesn't.

• Forecasting or time series in a messy operational setting.
• Recommenders or ranking systems with real feedback loops.
• Experience pairing classical models with LLMs in one product.

• Remote across MENA. Work from wherever you are in the region, with hours built around your team's overlap rather than someone's attendance spreadsheet.
• Work that ships. Products go live in weeks, not quarters. When someone asks what you have been building, you can open it on your phone and show them.
• Small teams, no ticket queue. Four to eight people carry a product from the first client call to handover. Nobody hands you a spec and disappears.
• Kit and tools you choose. A budget for your setup, and for the editor, licenses and services that make you fast. No three-week procurement saga.
• A learning budget with time attached. Money for the course or the conference, plus room in the week to use it. The second half is the part most places skip.
• Client contact from day one. You sit in the room where the decisions get made. Plenty of engineers wait a decade for that seat.
• Time off people take, and no timesheets. We measure engagements, not humans in fifteen-minute increments. Book the holiday, go, come back.