Careers in AI: A Roadmap for Students

Students often ask me: "Sir, how do I get a job in AI?" The field is broad, fast-moving and full of opportunity — in technology companies, research labs, startups, government, healthcare, finance, and in international, humanitarian and development organisations. This lecture offers a practical roadmap, based on what employers look for and what I have seen work for students.

The main roles#

You do not need to choose forever — careers evolve — but choosing a direction helps you focus.

A staged roadmap#

Stage 1 — Foundations (months 0–6)

• Python programming, Git, the command line.
• Mathematics: linear algebra, calculus, probability and statistics (the Mathematics track of this site).
• Data handling with pandas and SQL; visualisation.

Stage 2 — Core machine learning (months 4–10)

• Supervised and unsupervised learning with scikit-learn; evaluation and validation; feature engineering.
• Complete two or three end-to-end projects on real data.

Stage 3 — Deep learning (months 8–14)

• PyTorch; CNNs, sequence models, transformers; transfer learning and fine-tuning.
• One project in vision or NLP, ideally in a language or domain you know well.

Stage 4 — Specialise and ship (months 12+)

• Choose depth: NLP/LLMs, computer vision, MLOps, reinforcement learning, data engineering, responsible AI.
• Deploy something real: an API, a small web app, a dashboard used by actual people.
• Learn cloud basics, Docker and monitoring.

Throughout: read papers, write about what you learn, and practise communication.

Building a portfolio that stands out#

Employers see many portfolios with the same Titanic, MNIST and house-price projects. Stand out by:

1. Solving a real problem — ideally one from your community, workplace or field, with real (appropriately anonymised and permitted) data.
2. End-to-end work: data collection and cleaning → modelling → evaluation → deployment → monitoring and lessons learned.
3. Clear documentation: a README explaining the problem, approach, results with honest limitations, and how to run it; a model card for your model.
4. Rigour: proper validation, baselines, error analysis, subgroup evaluation.
5. Communication: a short blog post or video explaining the project to a non-technical audience.
6. Local-language or low-resource contributions: datasets, models or benchmarks for under-served languages are valuable and distinctive.
7. Open-source contributions: fixing documentation or bugs in libraries you use.

Strong project README outline
1. Problem & who it helps          5. Results (with baselines & confidence intervals)
2. Data (source, permission, ethics)   6. Error analysis & limitations
3. Approach & why                   7. How to run / demo link
4. Evaluation design                8. What I would do next

Finding opportunities#

• Internships and graduate programmes — apply broadly and early.
• Competitions (e.g. Kaggle) — good for practice; a strong finish is a credible signal.
• Research assistantships with faculty; co-authoring papers.
• Hackathons and open-source communities; local AI and data science meetups.
• International and humanitarian organisations: data, innovation and information-management roles, internships and fellowships; national and regional volunteering programmes.
• Freelancing on small projects to build experience.
• Networking with genuine curiosity: share your work, ask thoughtful questions, help others.

Interview preparation#

• Coding (data structures, Python, SQL).
• ML fundamentals: bias–variance, regularisation, evaluation metrics, overfitting, trees vs neural networks, how transformers work.
• Case studies: "How would you build a system to prioritise support requests?" — structure your answer: problem framing, data, baseline, model, evaluation, deployment, monitoring, risks.
• Your projects: be ready to explain every decision and what went wrong.

Growing through your career#

• Keep learning deliberately — the field changes quickly.
• Develop communication and collaboration; many projects fail from misunderstanding, not bad models.
• Seek mentors and, later, mentor others.
• Hold on to your values: the most respected AI professionals combine technical excellence with integrity about limitations and impact.