Allison Krinsky
Careers in AI are hot right now. And while many schools are adapting their curriculum to incorporate classes on the topic, there's still a short list of schools that offer an AI major.
Allison Krinsky graduated from the University of Washington with a degree in informatics in 2022. She now works as a data scientist at JPMorgan and makes videos about tech careers in her free time. She told Business Insider that many majors are interchangeable and that several degrees, such as computer science, math, information sciences, and data science, can lead to jobs in the field.
But even though Krinsky studied a traditional curriculum to get a tech job, she said her work at a research lab advanced her career more than anything else. She said during her year at the lab, she did a "heap" of things including building models and managing databases.
Most AI-related jobs require a technical portion in the interview process and Krinsky said candidates need to be able to talk about the projects they've done.
"A lot of times my interviews would just be people asking me about what I had built and what I did and the problems that I faced," Krinsky said.
Krinsky said while Big Tech names may look flashy on your résumé, hands-on experience is crucial to actually landing the job. In the internships she had before the research lab, she said she was given small projects that didn't involve too many skills.
"The internship is great to say somebody hired me, and that's a little bit of credibility," Krinsky said. "But you're not out of the game if you haven't had a traditional internship."
As AI jobs grow more in demand, some companies are growing increasingly picky about what they're looking for. So if you have limited experience or if you want to enhance your résumé, it's not a bad idea to build your own project and skill up. Krinsky said there are a number of avenues you could go down depending on the kind of roles you're interested in.
One option Krinsky recommends is a travel recommendation system built with large language models. She said you could do this project with limited experience and in different ways, like by using prompt engineering, retrieval augmented generation, or fine-tuning.
Krinsky also suggested creating a sentiment classification system for reviews, using natural language processing. She said this involves extracting information from text data and sorting it into entities like positive or negative sentiments. Krinsky said this can be used for financial analysis or identifying investment opportunities or risks.
Krinsky said you can also try an image recognition or computer vision project. This involves finding a set of pictures with labels and teaching a computer to identify what's in the images. She said it's a good way to learn about neural networks.
Krinsky said these projects can take between one and three months, depending on how much free time you have. Most projects start with scraping the web for data and then require building, training, and fine-tuning the model. Krinsky also recommended creating a report detailing the project process and results so that you have something to show for your work.
The projects don't have to be revolutionary, she said, but you should experiment with multiple data sets and be able to explain what's happening. She said anyone can recreate code from a tutorial so it's important to add a unique aspect.
"You have to get past 'I just wrote code and it didn't break,'" Krinsky said.
