AI work is much more than just writing code.
Just ask Pranjali Ajay Parth, a 25-year-old data scientist at Autodesk who develops AI tools that provide insights into employee work patterns, including meeting trends and work routines.
After earning her master's degree in computer science and working at Autodesk for more than a year, Peirce was able to understand what it's actually like to work in AI — and it's not what people expect, she said.
Peirce said that work in the AI field is largely interdisciplinary and relies on collaboration. She also said that even if you work in technology, you need to focus on ethics. In a conversation with Business Insider, she exposed some of the misconceptions about the role of AI.
It's not just about coding
Pranjali says if you're looking for a job in the AI field, being familiar with Python isn't enough.
Peirce said you don't necessarily need a degree in AI to land a job in the field, but you should know how to do case study analysis, SQL queries and coding, and candidates can take part in bootcamps and personal projects to hone their skills in these areas, he said.
“AI is inherently interdisciplinary,” Peirce says. “It's informed by many different disciplines: mathematics, computer science, statistics, and domain-specific knowledge.”
Peirce said about 70% of her job is data science, which involves reviewing and analyzing data sets. The rest of her time is spent on software engineering, building pipelines, data engineering, architecture design and a lot of math.
Peirce added that technology is constantly evolving, so it's important to stay up to date on developments in related fields.
AI roles are often highly collaborative
Software engineers are known to enjoy solitude, but don’t expect solitude if you work in AI.
While some engineering roles are independent, “AI projects are rarely done in isolation,” says Peirce, in part because AI is an emerging technology that requires collaboration across different teams and stakeholders, he says.
For example, Parse said building an AI recommendation system project required interacting with seven or eight teams.
In her experience, the process starts with data collection and preparation by the data analytics team. Then, data scientists apply statistical methods and modeling. Next, the machine learning team develops and refines the models. Once the models are ready, UX and UI experts design the user interface, followed by software engineers to build the front end.
Finally, the marketing team decided on the product launch strategy.
“End-to-end AI projects require a lot of communication and collaboration,” Peirce said.
We need to think about ethics
When sensitive data is handled during AI development, privacy teams are often heavily involved in the process.
According to Peirce, privacy protocols will be extensive: Employees will need to get permission to work with individuals' data. The project will also require robust production measures, such as pseudonymizing identities and ensuring models don't “incorrectly reproduce biases or produce unfair outcomes.”
This requires complying with legal and regulatory requirements, she said, and also means thinking about the long-term implications of the project, including potential unanticipated consequences or ethical dilemmas.
While privacy may seem like a natural consideration for those working in AI, it's easy to get caught up in model performance, Parse said, and with so many teams contributing to the product, it's easy to focus on specific tasks rather than the overall impact, she added.
Peirce said it's up to companies to educate their employees on proper privacy and ethical guidelines, but it's also important for employees to consider third-party perspectives on their work.
