“How many projects have you done so far?”
This question is common for interviewers to ask in data science. I’ve done interviews at my company for data analysts, data scientists, and data engineers. This question is quite often the jackpot of all questions. This one is especially true for a fresher or a relative newbie to the field.
Simply taking courses or attaining certifications isn’t enough. Almost everyone I talk to has certifications in various aspects of this field. It doesn’t add value to your resume by itself. You must combine certificates with hands-on experiences.
And that’s the value of open-source projects in this field. Data science projects are crucial. Interviewers love when applicants find projects and create solutions. This shows intellectual curiosity, enthusiasm, and passion for the field of data science. Trust me when I say to you that putting data science projects on your resume and portfolio will increase how employable you are.
Sources of Open-Source Projects
The next question is…“But…which data science projects should I do?” I like studying the best projects from previous months and attempting them. I intend to post my favorites as I find them.
In addition to this site, another good place to start looking at projects would be ‘Getting Started with GitHub’ if you’re new to GitHub. A third good place to check would be previous open-source data science projects, which has more than 100 projects.
DETR, An Example
An example of an open-source project for you to work on is DETR. You can attempt solving an image detection problem with DETR, which was made by the people of Facebook AI. DETR is one of the most interesting open-source projects. The fact that it amassed close to 3,000 stars in a single week is rather telling.
DETR, short for DEtection TRansformer, could be a game changer for the field of computer vision. This framework is innovative and efficient at dealing with challenges of object detection. Further, DETR is very fast and extremely efficient – it’s a dream for people in the field of data science.
“DETR doesn’t require installing a library to implement. DETR treats problems of object detection as direct set prediction problems via help from encoder-decoder architecture that is based on transformers.