Dhakaa Labs

Dhakaa builds applied machine learning for companies that already have the data and have not got anything out of it yet — Arabic document extraction, demand forecasting and computer vision on the factory floor. Every engagement starts with a two-week feasibility study that is allowed to end in no.

• 

  Feasibility study

  Two weeks against your own data, ending in a written answer with the accuracy we actually measured. It is allowed to say the model is not worth building, and about one in five does.

• 

  Arabic document extraction

  Invoices, contracts and government forms read into structured fields, in Arabic script and in mixed Arabic-English documents where most off-the-shelf tools quietly fail.

• 

  Demand & inventory forecasting

  A forecast built on your history and your calendar — Ramadan, Hajj season and school terms included, because a model trained on a generic year is wrong here twice a year.

• Computer vision on site

  Counting, defect detection and safety compliance from cameras you already have, running on a box in your rack rather than in someone else's cloud.

• MLOps & handover

  Deployment, monitoring and the retraining schedule, plus the training that lets your team run it after we leave. A model nobody can retrain has a shelf life.