Land Your First Job in AI, Machine Learning or Data Science
A 24-week plan for people who already code a little: pick one role, build three projects that prove it, make your CV, GitHub and LinkedIn work for you, reach the right people, and walk into the interview prepared.
Three working projects online, a CV and profiles that match one target role, 8 to 10 conversations with people already in the field, and 20 to 30 applications sent to roles you actually fit. Projects you can explain beat a list of certificates. Every phase below ends with a deliverable you can check off.
Before you start: you can write basic Python (variables, loops, functions) and you have a laptop and free GitHub, Kaggle and LinkedIn accounts. Starting from zero? Follow the 6-month AI roadmap (in Arabic) first, and see the best AI learning platforms (in Arabic) for courses. This guide picks up where learning ends: getting hired.
Pick one target role first
"AI job" covers very different work. Choose one role now; it decides which skills you polish and which projects you build. You can switch later, but a CV aimed at everything is read as aimed at nothing.
Answers business questions with data. Core skills: SQL, Python with pandas, statistics, charts and dashboards, explaining results to non-technical people. Best first role if you come from business, finance or operations.
Builds models that predict or explain. Analyst skills plus machine learning with scikit-learn, experiment design, evaluation metrics and feature engineering.
Puts models into production. Strong Python, PyTorch, APIs (FastAPI), Docker, cloud basics, monitoring and pipelines. Closest to software engineering.
Builds products on top of large language models. LLM APIs, retrieval (RAG) with embeddings and vector search, prompt design, evaluation of outputs, and agents that call tools.
The demand is real, and documented. The World Economic Forum's Future of Jobs Report 2025 (over 1,000 employers) lists big data specialists, fintech engineers and AI and machine learning specialists as the three fastest-growing jobs in percentage terms. The UAE National Strategy for Artificial Intelligence 2031 names "attract and train talent for future jobs enabled by AI" as one of its eight objectives. In the United States, the Bureau of Labor Statistics projects data scientist jobs to grow 35% from 2025 to 2035, with a median pay of $120,230 (May 2025). That figure is for the US only; there is no official equivalent for the Gulf, so check current postings for local pay.
Weeks 1 to 4: close your skill gaps
This is not a course marathon. The goal is to sit down on day one of a job and contribute. Write code every day on real data, not just watch videos.
Weeks 1 and 2 · Python for data
- Core Python: data structures, functions, classes, files, error handling.
- NumPy and pandas: arrays, DataFrames, cleaning, merging, groupby, missing values.
- Charts: Matplotlib and Seaborn.
- SQL: joins, aggregations and window functions. Most data roles use it every day.
- Deliverable: 5 small analyses on real Kaggle datasets. Each one loads, cleans, explores with at least 5 charts, and ends with a written summary. All of it on GitHub.
Weeks 3 and 4 · Machine learning you can explain
- Algorithms: linear and logistic regression, decision trees and random forests, support vector machines, k-means and PCA, gradient descent.
- Concepts: the bias-variance tradeoff, cross-validation, metrics (accuracy, precision, recall, F1, ROC AUC, confusion matrix), feature engineering and imbalanced data.
- Deliverable: a GitHub repo named
ml-from-scratch. Implement each algorithm with NumPy only, then rebuild it with scikit-learn and compare the results. Interviewers ask how these work; this repo lets you show it.
Weeks 5 to 10: build three projects that get you hired
Most people learn the theory and never ship anything. Three or four solid, deployed projects are your real CV. Pick datasets close to the sector you want to join (banking, health, government services, retail, energy).
Project 1 · End-to-end ML pipeline (weeks 5 and 6)
- Idea: churn prediction, credit risk scoring, price prediction or demand forecasting.
- Must include: a cleaning pipeline, exploratory analysis, justified feature choices, at least three models compared on clear metrics, hyperparameter tuning, and a simple app or API (Streamlit or FastAPI).
- README: problem, approach, results with numbers, screenshots, and how to run it.
Project 2 · Deep learning or computer vision (weeks 7 and 8)
- Idea: product defect detection, document classification, food recognition or medical image classification on a public dataset.
- Must include: PyTorch, transfer learning from a pre-trained model (for example ResNet, EfficientNet or ViT), data augmentation, training curves, per-class results and a confusion matrix, and a demo where someone uploads an image and gets a prediction.
Project 3 · LLM application (weeks 9 and 10)
- Idea: a question-answering app over a set of documents (RAG), for example policies or product manuals, ideally with Arabic documents.
- Must include: an LLM API (OpenAI, Anthropic or an open model), embeddings with a vector store (for example FAISS or Chroma), a small test set of questions with expected answers to measure quality, and a demo people can use.
- Also: enter at least 2 Kaggle competitions during these weeks and publish your notebooks. You do not need to win; you need to show how you work with messy data.
Where to host the demo for free. Streamlit Community Cloud deploys a Streamlit app straight from a GitHub repo at no cost. On Hugging Face Spaces, Gradio and Docker Spaces now need a paid plan, but a free personal account can still host up to 2 Gradio Spaces on ZeroGPU. Free apps go to sleep when nobody uses them, so wake yours up before an interview. Never put an API key in your code; use the platform's secrets setting.
Record a 2-minute demo video for each project, upload it to YouTube as unlisted, and link it at the top of the README. A reviewer who never runs your code will still see it work.
Weeks 11 to 14: make your CV, GitHub and LinkedIn work for you
GitHub
- Create your profile README: click + then New repository, name it exactly your GitHub username, select Public, turn on Add README, then Create repository. Write a short bio, your target role, your skills and links to your three projects.
- Pin your best work: open Your profile, click Customize your pins, choose up to 6 repositories, then Save pins.
- In every project repo: a README (problem, approach, results, how to run), clean commented code, and a
requirements.txtso anyone can reproduce it.
- Headline: role plus proof, not just a title. Example: "Junior ML Engineer | Computer vision and LLM apps | Python, PyTorch, SQL".
- About: your story in 4 to 6 lines, the problems you solve, and links to your projects. Put project links in Featured and list exact tools under Skills.
- Open to Work: click Me, open your profile, click Open to, then Finding a new job, fill in the roles and locations, and Save. Choose Recruiters only if you are employed (LinkedIn says it cannot guarantee complete privacy) or All LinkedIn Members, which also adds the #OpenToWork photo frame.
- Post 2 to 3 times a week: a project you built and what you learned, a concept explained simply, or your take on a new paper or release. Comment thoughtfully on posts from people at companies you want to join.
CV
- One page, one target role. Contact line with email, LinkedIn, GitHub and portfolio links.
- Skills as specific tools (Python, SQL, PyTorch, scikit-learn, Docker), not vague words.
- Each project bullet starts with a verb and ends with a measured result: "Built a churn model that reached 0.86 ROC AUC on 50,000 customer records" beats "Worked on churn prediction". Use your real numbers.
- Adjust the CV to each job description before you send it.
Weeks 15 to 18: reach the right people
Networking is not asking for a job. It is short, honest conversations with people doing the work you want. Message ML engineers and data scientists at your target companies (peers, not only managers), team leads, people who just posted about hiring, and alumni from your university.
Outreach message (edit every bracket)
Hi [Name],
I came across your [post / project / talk] on [topic] and found [one specific detail] really useful.
I am building my career in [data science / ML / AI engineering] and recently finished [your strongest project, one line]. I would value hearing about your experience at [Company] and any advice for someone entering the field.
Would you be open to a 15-minute chat in the coming weeks? Completely understand if you are busy.
Thank you,
[Your name]Questions for the call
- What does a typical week look like on your team?
- Which skills are hardest to find in candidates?
- What is the biggest problem your team is working on right now?
- If you were starting in AI today, what would you focus on?
- Is your team hiring, or expecting to grow?
Pace: 5 personal messages a week, aiming for 8 to 10 calls in this phase. Send a thank-you within 24 hours that mentions something specific they said, and connect on LinkedIn. A referral can come weeks later from one of these conversations.
Weeks 19 to 24: apply smart and prepare for interviews
Skip the 100 random applications. Pick 20 to 30 roles where you are a strong fit, and for each one:
- Map the job description's requirements to your projects, line by line.
- Tailor your CV so the most relevant project is first.
- Write a 3 to 4 sentence cover note linking one project to their needs.
- If you spoke with someone there, mention it and ask whether they would refer you.
UAE nationals: Nafis (nafis.gov.ae) is the federal programme that supports Emiratis working in the private sector, with employment and professional development programmes. Register there in addition to LinkedIn and company career pages.
Technical interviews for these roles usually combine some or all of the following rounds. Prepare for each:
Coding
Python problems on arrays, strings, hash maps, sorting and basic dynamic programming, plus SQL for data roles. Practise easy and medium problems on LeetCode; 50 or more is a solid base.
ML concepts
Explain the bias-variance tradeoff, L1 vs L2 regularisation, how a random forest limits overfitting, and how backpropagation works. Short, clear answers.
ML system design
Design a recommender, fraud detector or support chatbot: clarify the goal, choose metrics, describe the data pipeline, pick a model, then cover serving and monitoring.
Behavioural
Prepare STAR stories (Situation, Task, Action, Result): a hard technical problem, learning something fast, a team disagreement, and the project you are proudest of.
Deliverable: 20 to 30 targeted applications, 50 or more practice problems, at least 5 mock interviews with a friend or an online platform, and 3 to 4 system design scenarios practised out loud.
The 24 weeks on one page
Skills: 5 data analyses on GitHub, the ml-from-scratch repo, SQL practice.
Projects: ML pipeline, deep learning demo, LLM app, 2 Kaggle entries, a demo video for each.
Profile: GitHub profile README and pins, LinkedIn complete, 8 or more posts, CV ready.
Network: 20 or more personal messages sent, 8 to 10 calls held, thank-you notes sent.
Apply: 20 to 30 tailored applications, 50 or more practice problems, 5 mock interviews.
Resources worth your time
- Practical Deep Learning for Coders (fast.ai): free; uses PyTorch and Hugging Face. Expects about a year of coding experience.
- Machine Learning Specialization (DeepLearning.AI and Stanford Online, Coursera): 3 beginner courses by Andrew Ng. You can audit it, but the certificate is paid.
- Hugging Face LLM Course: free, the former NLP course, now focused on large language models.
- Kaggle Learn: short free courses with free certificates (Python, pandas, SQL, machine learning, feature engineering) plus guides on NLP, model evaluation and competitions.
- Made With ML (Goku Mohandas, now part of Anyscale): taking ML to production: testing, serving and CI/CD.
- Hugging Face Trending Papers: follow new research with code. It replaced Papers With Code, which now redirects there.
- Books: Hands-On Machine Learning with Scikit-Learn and PyTorch (Aurélien Géron, 2025), Designing Machine Learning Systems and AI Engineering (Chip Huyen), and Deep Learning (Goodfellow, Bengio and Courville, free to read online).
- YouTube: 3Blue1Brown (maths and neural networks), StatQuest with Josh Starmer (statistics and ML), and Andrej Karpathy's Zero to Hero series (former Director of AI at Tesla).
About certificates. The TensorFlow Developer Certificate that many older guides recommend is closed; Google has stopped the exam. Cloud certificates are aimed at experienced people: Google Cloud's Professional Machine Learning Engineer exam costs $200 and recommends 3 or more years of industry experience. For a first job, spend that time on projects.
Before you send a single application
Sources: World Economic Forum, Future of Jobs Report 2025; US Bureau of Labor Statistics, Occupational Outlook Handbook: Data Scientists; UAE National Strategy for Artificial Intelligence 2031 (ai.gov.ae); Nafis (nafis.gov.ae); GitHub Docs (profile README, pinned items); LinkedIn Help (Open to Work); Streamlit and Hugging Face Spaces documentation; TensorFlow certificate page; Google Cloud certification page; course pages of fast.ai, Coursera, Hugging Face, Kaggle and Made With ML. Checked November 2026.
Want your team to use this at work?
I'm Abdulaziz Alshamsi, a certified AI trainer and consultant in Abu Dhabi. I train government and company teams to use AI in their daily work. Message me to discuss a session for your team.