So much confusion, so many doubts and questions around careers in AI! When deciding a career pathway, the lack of clarity can be really stressful. Instead of continuing to answer individual questions, I decided to address them once and for all. Developing an AI project involves various technical roles and tasks. Let’s demystify them!
In this post we are going to go through the following:
- Roles and tasks in an AI project.
- Understanding the ins and outs of the tasks in the AI life cycle.
- Understanding the ins and outs of the roles in the AI life cycle.
- Career domains (who is hiring for AI roles, branches of AI, and future outlook).
- Learning resources.
1. Roles and Tasks in an AI Project
Developing an AI project is not a one person feat. It involves various distinct tasks and roles. This means that AI teams include individuals who focus on distinct parts of the product life cycle.
Here is a visual representation of the main technical roles and their relationship with the tasks in the development cycle:
Legend: 🟩 Primary tasks for that role. 🟨 Secondary tasks: things they do at times; don’t need to be proficient at them. 🟥 Tasks not needed for that role.
Please note that:
- The exact number of roles required to sustain a successful AI team vary based on the size and scale of the project. Not all the teams need to have people assigned to all the above mentioned technical roles. For example, very often AI teams do not need in house Researchers.
- Also, the exact boundaries between these roles and tasks can vary from company to company. For example, some teams might have Applied Scientists along with Data Scientists, while others might have ML Engineers along with Data Scientists; some might have both Data Analysts and Data Scientists, while others might just hire Data Scientists with strong business acumen and skip hiring Data Analysts. These boundaries tend to be blurry especially in startups.
But some basics first!
Before diving into further details, let’s quickly make sure that we have a clear picture of the fundamentals:
2. Ins and Outs of the Tasks in the AI life Cycle
i. Data Engineering
Task description: Prepare the data (input from a variety of databases) and transform it into a format that can be easily used by others in the AI life cycle, such as Analysts and Data Scientists.
Skills required: Coding and software engineering skills, with breadth varying based on the scale of the project.
Tools used: Database query languages such as SQL and object-oriented programming languages such as Python, C#, and Java. Big data tools such as Apache Spark are also commonly used.
Task description: Look for patterns in the data to help businesses with decision making and/or to automate processes, depending on the use case. For example, a real estate model could be used to predict selling prices of houses, while a computer vision model could be used for automatically detecting faulty items during manufacturing.
Skills required: Mathematics, data science, and machine learning.
Tools used: Python, TensorFlow, and PyTorch are commonly used.
Task description: Make the AI solution available to end users – combine the data with the model and put the end-to-end solution into production.
Skills required: Strong coding and back-end engineering skills to write production code that is robust and scalable.
Tools used: Object-oriented programming language such as Python, C#, and Java, and cloud technologies such as Azure and AWS.
iv. Business analysis
Task description: Evaluate the performance of the deployed models and their impact on business. For example, users’ click data could be used to evaluate the performance of a recommendation model and understand if it is providing the expected business value.
Skills required: Business acumen, data science for analytics, strong communication and data presentation skills.
Tools used: Presentation and analysis tools such as Excel, PowerPoint, Power BI and A/B testing software. Sometimes programming languages such as Python might be required.
v. AI Infrastructure
Task description: Build and maintain reliable, fast, secure, and scalable software infrastructure to enable all the components in the AI development cycle. For example, YouTube infrastructure needs to be able to handle watching of 5 billion videos every day in a seamless manner!
Skills required: Strong software engineering skills.
Tools used: Object-oriented programming language such as Python, Java, or C# and cloud technologies such as Azure and AWS.
Task description: Deep dive into specific research areas/problems with the goal of improving on the current state-of-the-art, stay updated about latest developments in the field, and publish papers.
Skills required: Strong analytical and critical thinking, writing skills, persistence, focus and many more!
Tools used: Channels such as academic publications, Arxiv, Twitter and conferences such as NeurIPS, ICLR, and ACM for staying current. Python, PyTorch and/or TensorFlow for modeling.
3. Ins and Outs of the Roles in the AI Life Cycle
i. Software Engineer
This role requires working on data engineering and core infrastructure work. This means that strong coding and software engineering skills such as knowledge of algorithms and distributed systems is required.
ii. Data Scientist
This role requires working on data engineering, modeling, and business analysis tasks. The level of skills required, e.g., breadth and depth of engineering and business acumen required vary based on the overall team composition (description of the next few roles will make this clearer, especially see “Applied Scientist vs Data Scientist”) and scale of the project.
iii. ML/AI Engineer
This role requires carrying out data engineering, modeling, deployment and AI
infrastructure tasks. This is a versatile role which could be viewed as a combination of Software Engineer + Data Scientist, with stronger engineering skills than data science skills.
iv. Applied Scientist
This role is another variation of Software Engineer + Data Science combination.
ML/AI Engineer vs Applied Scientist: Compared to ML/AI Engineers, Applied Scientists are more heavy on data science skills than on software engineering.
Data Scientist vs Applied Scientist: In teams that have both Data Scientists and Applied Scientists, Applied Scientists tend to deal with more advanced ML concepts (deep learning), while Data Scientists are likely to be people with stronger business acumen.
v. Data Analyst
This role is about working on data engineering and business analysis tasks. However, unlike the other roles dealing with data engineering, data analysts don’t need proper coding and software engineering skills – their tasks can be accomplished using query languages such as SQL and using interactive tools such as Power BI and Excel (see “skills required and tools used” under “Business Analysis” task for details).
vi. AI/ML Researcher
Researchers are heavily focused on modeling using their strong scientific skills. They work on improving existing state-of-the-art models. In an industry setting, they provide guidance to others in the AI development cycle, such as Data and Applied Scientists, on how to improve current solutions and innovate using AI.
4. Career Domains
a. Who is Hiring for AI Roles
Tech giants like Microsoft, Google, Amazon, Facebook, Netflix, IBM, Uber and co are the best places to work in this field because besides working on cutting edge technology, you get to work on products that directly impact billions of end users:
Cortana, Alexa, Google Translator, Power Point Designer, Smart Reply and Smart Compose in Outlook and Gmail, YouTube and Netflix Recommendations, Face Recognition, People You May Know, Pages You Might Like, Food Delivery Estimates, Friends Suggestions, Feed Ranking, and the list goes on and on.
However, the applications of AI are so diverse that almost every industry is now embracing AI and hiring for talent in their vertical. Let’s look at some less obvious domains and their uses of AI:
- Finance: Fraud detection, prediction and execution of trades at speeds and volumes that humans can’t compete with.
- Medicine: Medical diagnosis, drug discovery, understanding of risk factors for diseases in large populations, expert systems to aid GPs, monitoring, and control in intensive care units, prosthetics design.
- Robotics: Vision control, motor control, learning, cooperative behavior.
- Engineering: Fault diagnosis, predictive maintenance, intelligent control systems, intelligent manufacturing systems.
- Marketing: More targeted, relevant, and timely marketing programs.
- Online customer support: Chatbots to replace customer support representatives.
- Space: NASA uses AI to help plan and schedule space shuttle maintenance.
- Military Activities: Huge influx of funds in here, but details are classified information!
b. Branches of AI
As we can see from the examples above, AI applications can be very diverse. To better illustrate this, here is an overview of the branches of AI:
c. Future Outlook
Also, here are some great insights from McKinsey Research for you to know which industries to keep an eye on:
5. Learning Resources
At this point you should have been able to narrow down the tools and skills for your desired role. From here onward, you should be able to find lots of relevant resources such as:
- Courses, bootcamps and certifications
a. Book Recommendations
- Machine Learning: Hands-On Machine Learning with Scikit-Learn and TensorFlow: Concepts, Tools, and Techniques to Build Intelligent Systems
- Statistics: Practical Statistics for Data Scientists: 50 Essential Concepts
- Data Science: Python Data Science Handbook
- Business Analysis: Storytelling with Data: A Data Visualization Guide for Business Professionals
b. Online Courses
Here are some great online courses and specializations:
- For deep dives into machine learning and deep learning, I highly recommend Andrew Ng’s (professor at Stanford) courses on Coursera:
- Besides Coursera, fast.ai is also a great to place to learn about deep learning.
- For learning Data Science in Python in a fun and interactive way, here is a highly loved course by yours truly!:)
Needless to say that there are tons of resources out there. If you want me to provide further recommendations (books/blogs/courses) around some particular area, please let me know in the comments.
AI is a relatively new and rapidly evolving field, so understandably there is a lot of confusion around. I hope you found this concise yet comprehensive guide helpful and you are walking away with a clearer picture regarding the AI pathway you want to pursue. Again, please read the job description and do a little research about the company/team you are applying for in order to understand which of the role(s) their description maps to — don’t rely on the title alone.
This kind of breakdown is not easy to find, so if you found this post insightful, kindly share it with your network to help me reach out to others who would benefit from it as well. Thanks!
Last but not the least, please remember that no matter which path you end up choosing, the most important thing is that your career choice should reflect your values.
Best of luck with the journey ahead!
. . .