Pranjali Ajay Parse
There's more to an AI job than simply writing code.
Just ask 25-year-old Pranjali Ajay Parse, who works as a data scientist for Autodesk. She's been developing an AI tool that provides employees with insights into their work patterns, such as meeting trends and work routines.
After getting her master's degree in computer science and working at Autodesk for over a year, Parse has been able to grasp what it's like to actually work in an AI role — and she said it's not what people may expect.
Parse said that working in AI is largely interdisciplinary and dependent on collaboration; and while you may be working in tech, the job also requires a heavy focus on ethics. In a conversation with Business Insider, she debunked some of the myths about AI roles.
Pranjali said proficiency in Python won't cut it if you're looking for a job in AI.
Parse said candidates don't necessarily need a degree in AI to get a job in the field. But she said you need to know how to do case study analysis, SQL querying, and coding. She said candidates can try boot camps or personal projects to skill up in those areas.
"AI is inherently interdisciplinary," Parse said. "It draws from various domains, including mathematics, computer science, statistics, and domain-specific knowledge."
Parse said about 70% of her job is data science, which requires reviewing and analyzing data sets. She said the rest of her time is split between software engineering, building pipelines, data engineering, architectural design, and a lot of math.
Parse also added that it's important to stay updated with advancements in related fields because the technology is constantly evolving.
Software engineers have been known to be loners, but don't count on solitude if you're working in AI.
While some engineering roles tend to be independent, Parse said, "AI projects are rarely done solo." Part of this is because AI is a new technology that requires collaboration among a variety of teams and stakeholders, she said.
For example, Parse said she has to interact with seven or eight teams to build an AI recommendation system project.
In her experience, the process begins with data collection and preparation by a data analysis team. Then, data scientists apply statistical methods and modeling. The machine learning team then develops and refines the model. Once the model is ready, UX and UI experts design the user interface, followed by software engineers who build the front end.
Finally, the marketing team determined the product's launch strategy.
"An end-to-end AI project requires a lot of communication and collaboration," Parse said.
Privacy teams are often deeply embedded in the process when sensitive data is handled during AI development.
Parse said the privacy protocols are extensive. When working with a person's data, employees need to receive permission for tasks. Projects also require robust production measures, like pseudonymizing identities and ensuring models don't "inadvertently recreate biases or create inequitable outcomes."
This requires complying with legal and regulatory requirements, she said. It also means thinking about the long-term implications of projects, including potential unintended consequences and ethical dilemmas.
While privacy may seem like an obvious consideration for those working in AI, Parse said it can be easy to get caught up in how the models perform. Also, since so many teams contribute to the product, it can be easy to focus on your specific task rather than the overarching implications, she added.
Parse said it's up to companies to train employees on proper privacy and ethical guidelines. But it's also important for employees to consider a third-person perspective on the work they're doing.
