In September 2025 we ran our first Data Science & Machine Learning Bootcamp, a one-month, hands-on program built around a simple idea: take a complete beginner through the entire ML workflow, from raw data to a deployed model, in real time. It was our first cohort, hosted alongside Sharafdin Yusuf and sponsored by Dugsiiye, and it became the template every bootcamp we've run since has built on.
Why we started
We'd spent time building datasets, tokenizers, and models for Somali NLP, and kept running into the same limiting factor: there weren't enough Somali engineers who felt confident in data science and machine learning to contribute to that work. Tutorials and courses existed, but almost none of them were built with Somali learners, or Somali context, in mind. So we built one.
The goal was deliberately narrow: not a semester-long computer science curriculum, but a focused, one-month sprint that gets someone from zero to a working, deployed project.
The one-month format
Every lesson maps to one stage of the machine learning workflow: collect data, preprocess it, split it into train and test sets, choose a model, train it, evaluate it, and deploy it. Rather than teaching those steps as abstract theory, each stage got its own lesson, its own assignment, and (for the harder stages, like regression, classification, and clustering) a dedicated "in action" walkthrough where we worked through the code live.
All lessons were recorded and published as unlisted videos on the Somast YouTube channel, so the material could be followed on a participant's own schedule, then paired with an assignment due before the next session.
What the curriculum covers
- An introduction to machine learning and how the end-to-end workflow fits together
- Data foundations: cleaning, exploring, and preparing a dataset before any model touches it
- A dedicated "Data Processing in Action" session, run live over Zoom
- Regression, taught in both theory and a hands-on Jupyter Notebook walkthrough
- Classification, again paired with an in-action coding session
- Clustering, covering unsupervised approaches to grouping data
- Deployment, closing the loop from trained model to something a user can access
- A final project pulling every stage of the workflow into one submission
We also added two short resource videos, installing Anaconda and setting up Jupyter Notebook, after realizing that environment setup was, for a real share of participants, the first real obstacle standing between them and lesson one.
Learning by shipping
The structure was intentional: watch a lesson, do the assignment, then move on. By the time participants reached the final project, they weren't seeing the ML workflow for the first time, they'd already collected, cleaned, modeled, and evaluated data across several earlier assignments. The final project just asked them to do it once more, on their own, start to finish.
Open from day one
Every lesson, assignment, and the final project brief live in the public sep-ds-ml-bootcamp-2025 repository on GitHub, released under a Creative Commons BY-NC-SA 4.0 license. That was a deliberate choice: the bootcamp doesn't disappear once the cohort ends. Anyone can clone the repo, follow the videos in order, and work through the same material independently.
The bootcamp was fully sponsored by Dugsiiye, which meant every seat was free for participants, the only cost was the time and effort to show up and do the work.
What this set in motion
This cohort was our first, and it taught us as much as it taught the participants. We learned which lessons needed a live walkthrough and which worked fine as a recorded video, where environment setup broke people's momentum, and where the pacing needed to change. Every one of those lessons shows up in the cohorts that followed: a second Data Science & Machine Learning cohort, a dedicated Python foundations bootcamp for people who needed programming fluency first, a third and more expansive DS/ML cohort, and eventually a Git & GitHub bootcamp for the collaboration skills the projects themselves depend on.
We're covering each of those in their own posts. This one is where the series starts.
Writing about Somali language technology, open data, and AI from the lab in Mogadishu.