AI Startup Newbie: How Should You Prepare Your Portfolio?
I'm a job seeker preparing for an AI engineer position. I've done a few deep learning projects during my undergraduate studies, but I'm not sure if they meet the standards that actual startups are looking for. I've uploaded model training code and a simple demo app on GitHub, but I've been advised to emphasize 'business impact' in interviews. For those currently working in the field, what do you focus on most in a portfolio? Would self-driven projects be better than Kaggle?
6 answers
I've had the same concerns, but I've come to feel that the process of defining and solving a problem is more important than Kaggle rankings. You're doing great by even uploading a demo app to GitHub! Additionally, it would be good to add explanations in the README for each project about 'why this model is needed' and 'what data it was trained on.' The key, I've found, is being able to naturally tell that story in an interview.
Great question! As a current professional, I prefer self-driven projects. Kaggle data is already cleaned, so it's far from real work. Instead, try calculating 'cost-effectiveness' in your projects. For example, even an estimate of how much an image classification model reduced manual inspection time would be impressive. Since you've uploaded it to GitHub, how about adding a demo video?
I think Kaggle isn't a bad option either. When you lack your own projects, you can appeal by achieving top rankings on Kaggle or highlighting original Feature Engineering work. However, for startup interviews, you need to practice connecting 'why you did this and how you reduced costs.' Good luck!
Have you ever integrated a project with actual user feedback? Startups value rapid iteration, so experience building an MVP and improving it is a big plus. For example, if it's a chatbot project, try including a case in your portfolio where you deployed it, analyzed logs, and improved performance. I believe business impact comes from that kind of 'real-world application'.
I agree! I had similar concerns when I was a junior, and I found it helpful to showcase the 'full pipeline from data collection to deployment' in my portfolio. For example, even a simple API with a model hosted on AWS or GCP can be a great talking point in interviews. As for business impact, even with small projects, you can naturally highlight it by specifying 'where this could be applied.'
I'm a current AI startup engineer. Business impact shows what real problem a project solved and how the results contributed to revenue or efficiency. For example, emphasize concrete numbers like '30% reduction in customer response time' rather than '95% model accuracy.' Personal projects are far better than Kaggle. Startups value hands-on experience much more.