The NYSC year is an odd gap in many graduates' lives. You have a degree, a posting, a monthly allowance and, depending on your place of primary assignment, some free evenings and weekends. For a growing number of corps members that gap is also a chance to change direction towards artificial intelligence. One year will not make anyone a senior engineer, but it is enough, with steady effort, to become a capable junior who can show real work. Here is a practical plan.
Be honest about the starting point
Begin by taking stock. What did your degree give you? Graduates of computer science, engineering, mathematics, physics or statistics usually have some programming and enough mathematics to start quickly. Graduates of other disciplines can still succeed, but should budget extra time for mathematics and coding basics, and should treat their own field as an asset. An agriculture or pharmacy graduate who understands machine learning is more useful in those industries than a generalist.
Then look at your practical situation. You will need a laptop: 8 GB of memory is workable, 16 GB is more comfortable. You will need data and power. Free cloud notebooks such as Google Colab and Kaggle let you train models on remote hardware, which spares your laptop, but they still need a connection. Work out what a month of data and backup power will cost where you are posted, and put it in your budget alongside transport and feeding.
Finally, be realistic about time. Ten to fifteen focused hours a week, kept up for a year, will take you much further than bursts of enthusiasm followed by weeks of nothing. Look at your PPA hours and your CDS day and decide which evenings and which part of the weekend belong to study.
Months one to three, the foundations
The first quarter is about tools and basics. Resist the urge to jump straight to chatbots and image generators; without foundations you will copy code you do not understand.
Focus on:
- Python, including functions, files and working with libraries
- NumPy and pandas for handling data
- the essential mathematics: linear algebra, probability and basic calculus
- Git and GitHub, used from the first week
- SQL, because most real data lives in databases
- plotting, so you can see what your data is doing
Free university course materials and official documentation are plentiful. The important habit is to type and run everything, then change it and see what breaks. By the end of month three you should be able to load a messy spreadsheet, clean it, summarise it and explain what you found.
Months four to eight, real projects
The middle of the year is where you become an engineer rather than a student. Learn classical machine learning with scikit-learn: regression, classification, decision trees, evaluation and the danger of overfitting. Then move to deep learning with a framework such as PyTorch, starting with simple image and text models.
Learning without building does not stick, so attach each topic to a project. The best projects use data close to home, because they force you to handle real problems and because they make your portfolio memorable. Some ideas:
- predicting food prices in local markets from figures you collect over several weeks
- classifying crop diseases from photographs taken on a phone
- translating or classifying text in Yoruba, Hausa or Igbo
- forecasting power supply in your area from your own outage log
For each project, write a short report covering the problem, the data, what you tried, what failed and what you would do next. Employers read these far more closely than certificates, because they show how you think.
By month eight, aim to have taken at least one project all the way to something another person can use, such as a simple web page or API that accepts input and returns a prediction. Deployment is where many self-taught learners stop, and it is exactly what employers want to see.
Months nine to twelve, depth and visibility
In the final quarter, choose a direction and go deeper. Options include computer vision, natural language processing, data engineering, or building large language models into real software. Pick one that matches the projects you enjoyed and the jobs you have seen advertised.
At the same time, make your work visible. Tidy your GitHub, pin your three best projects, and write a short article about each. Join AI communities and meetups in Lagos, Abuja, Ibadan and online, and share what you are building; many first jobs come from someone who has seen your work. Kaggle competitions are a good test of skill and give you a public record.
Start applying from month ten, not after passing out. Suitable first roles include junior machine learning engineer, data analyst with a machine learning focus, AI support engineer and research assistant. Internships, even short ones, are worth taking. Remote roles with foreign employers are possible but competitive, so your portfolio has to do the talking.
Some people prefer structure to self-direction. A full-time or weekend programme gives you a syllabus, tutors and deadlines, which suits many graduates. RAIN's AIML programme, for example, runs for 12 months, physical or virtual, which fits naturally alongside or just after a service year. Whether you choose a programme or go it alone, the plan above still describes what you need to have done by the end.
Keeping going when it is hard
There will be weeks when primary assignment duties take over, the light is gone for days, or an idea refuses to make sense. That is normal. Keep a simple log of what you studied each week; looking back at it on a bad day shows how far you have come. Find one or two study partners, even online, and agree to show each other something every week.
Do not compare your month three with someone else's year five on social media. Compare yourself with where you were when you started.
A year from now you can be a graduate who has served, who has a portfolio of real projects, and who can explain clearly how each one was built. That person gets hired.
Start this week by fixing your study hours, opening a GitHub account and finishing your first small Python exercise.