Picture by Creator
Vibe coding is likely one of the largest AI tendencies in 2025.
In case you haven’t already heard about it, vibe coding is basically AI-assisted coding. You simply describe what you wish to construct, and AI creates the whole software for you. In case you run into errors, the mannequin will repair them. It generates, assessments, and debugs code with restricted human intervention.
Whereas many programmers have raised considerations about this development, calling it a “dangerous shortcut” and mentioning legitimate dangers like elevated long-term technical debt, the “vibe coding” development isn’t going away anytime quickly.
And whereas I’ve seen a ton of hype round vibe coding within the software program growth discipline, I haven’t heard many information scientists speak about it.
As a knowledge scientist myself, I “vibe code” virtually day by day at work. If used correctly, I believe this method may also help you turn into a extra environment friendly information scientist. Actually, in case you are an aspiring information scientist, studying the appropriate AI instruments will assist you to keep forward of the competitors. You may convey your concepts to life by constructing inventive portfolio tasks — serving to you stand out to potential employers. Alongside the best way, additionally, you will find out about new frameworks that can assist you to turn into a greater information scientist.
On this article, I’ll clarify how one can construct information science tasks with vibe coding.
Examples of Vibe Coding Knowledge Science Tasks
Once I first began out within the discipline of knowledge science, I constructed many portfolio tasks and deployed them. Right here is an instance of a facial recognition app I created years in the past:
Picture by Creator
I constructed a neural community that might predict an individual’s celeb lookalike from a photograph they uploaded. This was one of many many portfolio tasks that helped me land my first information science job.
Probably the most time-consuming a part of this challenge? Deployment.
It took me days to create the front-end net software — designing the app, permitting customers to add a photograph of themselves, constructing the progress bar, and producing the output.
You see, like many different information scientists, I don’t know front-end programming languages like HTML and CSS. My focus is totally on constructing fashions and analyzing information with Python. Nevertheless, employers are now not impressed by the Python code in your GitHub repository.
A front-end software just like the one I showcased above is way extra interesting, because it permits hiring managers and employers to work together with the fashions you will have constructed. For aspiring information scientists, that is the place vibe coding is available in. In simply minutes, you may construct a machine-learning mannequin and get AI to develop an entire net software just like the one above.
Listed below are two purposes I created with AI in lower than 5 minutes:
1. Twitter Sentiment Evaluation Mannequin
Picture by Creator
2. Titanic Survival Prediction Mannequin
Picture by Creator
Let’s discover how one can construct tasks just like the above to impress potential employers and get employed as a knowledge scientist.
Constructing Knowledge Science Tasks with Vibe Coding
Step 1: Selecting an AI Instrument
You may simply vibe code with instruments like ChatGPT and Claude by explaining your necessities to the AI and pasting its output into your programming IDE.
Nevertheless, I counsel going a step additional and utilizing an AI assistant that integrates immediately along with your IDE, comparable to Cursor AI, Lovable, and Copilot in Visible Studio Code.
These instruments will entry your complete codebase (datasets or every other information in your listing). They may then analyze your codebase, generate, and immediately run code that meets your necessities.
I take advantage of Cursor AI at work virtually day by day, and it saves me a ton of time.
To get began with Cursor AI, you may go to this web site, comply with the set up directions, and have it arrange in seconds. You’ll then see a display screen that appears like this with a chat field on the appropriate, the place you may kind prompts.
Picture by Creator
In case you’d like a extra complete information on getting began with Cursor AI, I like to recommend watching this video.
Step 2: Producing Venture Concepts
The subsequent step is to ideate. You wish to create a challenge that’s distinctive — one thing that potential employers will discover fascinating.
If you end up caught, you may kind the next immediate into an AI chat interface to get some concepts:
Generate 5 distinctive information science portfolio challenge concepts that embrace each analytical parts and interactive front-end dashboards. These tasks ought to use non-traditional datasets.
Step 3: Making ready your Codebase
In case you’re utilizing an AI assistant like Cursor that has entry to your codebase, then create a listing with all of the challenge information and datasets required. For instance, if you wish to construct a sentiment evaluation mannequin, your listing will comprise a single coaching dataset.
In case your challenge is extra advanced, I counsel making a separate file to doc the next:
- A markdown file containing your challenge’s necessities. If in case you have a selected request, like guaranteeing the mannequin avoids utilizing a library that is perhaps deprecated, you may add it right here.
- For domain-specific code or lesser-known libraries, embrace documentation so the mannequin doesn’t run into too many errors.
- Examples of code notebooks with related workflows (this protects time as a result of it offers the mannequin context and tells it strategy the issue).
Step 4: Creating the Immediate
When you’ve ready your listing with related information, you can begin prompting the AI to provide you code to construct the challenge.
In Cursor, there’s a chatbox on the right-hand aspect of the display screen that permits you to do that.
Listed below are some prompting greatest practices when vibe coding:
- Present a ton of context: To keep away from working into errors, you will need to present the mannequin with as a lot context as attainable. With instruments like Cursor, you may even embrace a picture of what you’d like the tip product to appear to be.
- Pressure it to learn your paperwork: When working with giant codebases, I’ve observed that AI assistants are inclined to skip over the information I uploaded and as an alternative begin to hallucinate column names. To forestall this from taking place, it’s essential explicitly immediate the AI to learn by means of particular paperwork earlier than producing code.
- Function-playing: You can too ask AI to tackle the position of a site knowledgeable earlier than constructing a challenge. This strategy may also help you generate richer output because it tells the mannequin to attract from a subset of its information base that pertains to a selected area.
Right here is an instance of a immediate I used to construct a sentiment evaluation net software with Cursor:
Create a sentiment evaluation net app that:
1. Makes use of a pre-trained DistilBERT mannequin to investigate the sentiment of textual content (constructive, unfavourable, or impartial)
2. Has a easy net interface the place customers can enter textual content and see outcomes
3. Exhibits the sentiment outcomes with applicable colours (inexperienced for constructive, pink for unfavourable)
4. Runs instantly without having any coachingPlease join all of the information correctly so after I enter textual content and click on analyze, it exhibits me the sentiment outcomes instantly.
Step 5: Iterate and Refine
AI fashions like Cursor can deal with smaller tasks simply high-quality however are inclined to hallucinate and run into errors when working with bigger codebases.
That is the place your area experience is available in; as an example, explaining to the AI how a sure metric ought to be outlined, which dataset to tug from, and even telling it which libraries to make use of and what to keep away from.
One strategy I’ve discovered particularly helpful is switching between a device like Cursor and a special LLM like Gemini 2.5 Professional. In case you discover Cursor working into the identical error a number of instances, it is perhaps since you aren’t explaining your necessities comprehensively. You may copy and paste the error into a special LLM and get it to generate a complete immediate, which you’ll then paste into Cursor.
This multi-LLM strategy works properly after I run into errors whereas vibe coding.
Vibe Coding for Knowledge Science: Yay or Nay?
I personally don’t consider which you can vibe code your strategy to production-ready code.
If you’re a knowledge scientist, I nonetheless counsel studying Python and SQL; in any other case, you’ll find yourself with computationally costly code that results in long-term technical debt.
Nevertheless, I believe vibe coding has its deserves for particular use instances, comparable to constructing a knowledge science portfolio challenge you probably have no front-end coding information.
You can too use it to hurry up your workflows and rapidly be taught new libraries and methods you aren’t acquainted with — so long as you keep in mind core programming ideas and don’t get over-reliant on AI.
As a subsequent step, I like to recommend familiarizing your self with the next AI instruments that can make you a extra environment friendly information scientist:
 
 
Natassha Selvaraj is a self-taught information scientist with a ardour for writing. Natassha writes on the whole lot information science-related, a real grasp of all information subjects. You may join along with her on LinkedIn or try her YouTube channel.
