Summary
There are 3 steps to career growth: learning (technical and misc. skills), working on projects and searching for a job. In AI, the foundation are still maturing—changing—and you must keep up to date with them. You will also have to educated your colleagues and interviewers about AI such that they can understanding the reasoning behind decisions.
The most important topics for a technical career in AI are:
- Foundational Machine Learning skills. Commonly used models (linear regression, logistic regression, gradient descent, neural networks, decision trees, clustering and anomaly detection), how they work and core concepts behind how and why machine learning works (bias/variance, cost functions, regularization, optimization algorithms and error analysis).
- Deep Learning. The basics of neural networks, how to make them work (e.g. hyperparameter tuning), convolutional networks, sequence models and transformers.
- Math relevant to Machine Learning. Linear algebra (vectors, matrices and how to manipulate them), probability & statistics (discrete and continuous probability, standard probability distributions, basic rules such as independence and Bayes rule, and hypothesis testing), exploratory data analysis (EDA) and calculus.
- Software development. Programming fundamentals, data structures (especially those that relate to machine learning, such as data frames), algorithms (including those related to databases and data manipulation), software design, Python, key Python libraries (TensorFlow, PyTorch, scikit-learn).
When learning you should follow a course so that there is consistency between topics. After courses, you can switch over to research papers and other resources. Consistent learning over a long period of time makes you capable.
Once you have enough knowledge to create, start by working on smaller projects and ‘graduate’ to larger ones as you showcase capability. If you don’t have a project idea of your own, you can join existing projects. Given numerous options, decide on a focus by analyzing the potential technical growth, teammates and potential professional growth.
When beginning a project there are 2 approaches to consider: ‘Ready, Aim, Fire’ and ‘Ready, Fire, Aim’. The best approach is situation dependent. ‘Ready, Aim, Fire’ requires careful planning and due diligence. You should only commit and execute when there is a high degree of confidence. This is best when the cost of execution is high and mistakes costly. ‘Ready, Fire, Aim’ is the process of jumping into development and execution. This allows for the rapid discovery of issues and ease in pivoting. This is best when the cost of execution is low, allowing you to find suitability—or lack of—through trial and error. With Machine Learning model development being an iterative process, the ‘Ready, Fire, Aim’ process generally has more merits and can lead to better solutions, after a product direction has been decided upon. This process of iteration makes sense when there are multiple options available (all of which are potentially viable) and where experiments can be ran (and therefore data on performance acquired) quickly. This is true for data labeling (Creating labeling guidelines that result in clean and consistent labels), model training (Selecting data, hyper-parameters and architecture) and deployment & monitoring (Finding the most valuable metrics).
When applying for a job, a resume should answer the question ‘Why should I select you?’. This can be more effectively answered if the resume if tailored specifically for the role and company. You should explain how you have used skills instead of just listing them as this showcases knowledge instead of potentially listing buzzword terms. Include resulting metrics that show the impact of your performance. Provide accessible links where necessary. By working on personal projects, you demonstrate a working knowledge and interest, and that you’re staying up to date with the field. In the case of rejection, find out why you were rejected and work to improve problematic areas.