Picture by Creator
Vibe coding is without doubt one of the greatest AI traits in 2025.
If you happen to 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. If you happen to run into errors, the mannequin will repair them. It generates, exams, and debugs code with restricted human intervention.
Whereas many programmers have raised considerations about this pattern, calling it a “dangerous shortcut” and stating legitimate dangers like elevated long-term technical debt, the “vibe coding” pattern isn’t going away anytime quickly.
And whereas I’ve seen a ton of hype round vibe coding within the software program growth subject, I haven’t heard many knowledge scientists speak about it.
As an information scientist myself, I “vibe code” nearly each day at work. If used correctly, I feel this method can assist you change into a extra environment friendly knowledge scientist. In actual fact, in case you are an aspiring knowledge scientist, studying the best AI instruments will assist you to keep forward of the competitors. You possibly can convey your concepts to life by constructing inventive portfolio initiatives — serving to you stand out to potential employers. Alongside the best way, additionally, you will study new frameworks that may assist you to change into a greater knowledge scientist.
On this article, I’ll clarify how one can construct knowledge science initiatives with vibe coding.
Examples of Vibe Coding Information Science Tasks
Once I first began out within the subject of information science, I constructed many portfolio initiatives 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 initiatives that helped me land my first knowledge science job.
Probably the most time-consuming a part of this mission? Deployment.
It took me days to create the front-end internet 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 knowledge scientists, I don’t know front-end programming languages like HTML and CSS. My focus is totally on constructing fashions and analyzing knowledge 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 much extra interesting, because it permits hiring managers and employers to work together with the fashions you will have constructed. For aspiring knowledge 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 internet 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 initiatives just like the above to impress potential employers and get employed as an information scientist.
Constructing Information Science Tasks with Vibe Coding
Step 1: Selecting an AI Instrument
You possibly can 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 recommend 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 whole codebase (datasets or another information in your listing). They are going to then analyze your codebase, generate, and immediately run code that meets your necessities.
I exploit Cursor AI at work nearly each day, and it saves me a ton of time.
To get began with Cursor AI, you may go to this web site, observe the set up directions, and have it arrange in seconds. You’ll then see a display that appears like this with a chat field on the best, the place you may kind prompts.
Picture by Creator
If you happen to’d like a extra complete information on getting began with Cursor AI, I like to recommend watching this video.
Step 2: Producing Undertaking Concepts
The subsequent step is to ideate. You wish to create a mission that’s distinctive — one thing that potential employers will discover attention-grabbing.
If you end up caught, you may kind the next immediate into an AI chat interface to get some concepts:
Generate 5 distinctive knowledge science portfolio mission concepts that embrace each analytical elements and interactive front-end dashboards. These initiatives ought to use non-traditional datasets.
Step 3: Making ready your Codebase
If you happen to’re utilizing an AI assistant like Cursor that has entry to your codebase, then create a listing with all of the mission 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 mission is extra complicated, I recommend making a separate file to doc the next:
- A markdown file containing your mission’s necessities. You probably 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 comparable workflows (this protects time as a result of it provides the mannequin context and tells it find out how to strategy the issue).
Step 4: Creating the Immediate
When you’ve ready your listing with related information, you can begin prompting the AI to offer you code to construct the mission.
In Cursor, there’s a chatbox on the right-hand facet of the display that means that you can 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 top product to appear like.
- Power it to learn your paperwork: When working with massive codebases, I’ve observed that AI assistants are likely to skip over the information I uploaded and as an alternative begin to hallucinate column names. To stop this from occurring, it is advisable to explicitly immediate the AI to learn via particular paperwork earlier than producing code.
- Function-playing: You too can ask AI to tackle the function of a site professional earlier than constructing a mission. This strategy can assist 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 internet software with Cursor:
Create a sentiment evaluation internet app that:
1. Makes use of a pre-trained DistilBERT mannequin to investigate the sentiment of textual content (constructive, adverse, or impartial)
2. Has a easy internet interface the place customers can enter textual content and see outcomes
3. Reveals the sentiment outcomes with applicable colours (inexperienced for constructive, purple for adverse)
4. Runs instantly without having any coachingPlease join all of the information correctly so once I enter textual content and click on analyze, it exhibits me the sentiment outcomes straight away.
Step 5: Iterate and Refine
AI fashions like Cursor can deal with smaller initiatives simply effective however are likely to hallucinate and run into errors when working with bigger codebases.
That is the place your area experience is available in; for 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 software like Cursor and a unique LLM like Gemini 2.5 Professional. If you happen to discover Cursor working into the identical error a number of occasions, it is perhaps since you aren’t explaining your necessities comprehensively. You possibly can copy and paste the error into a unique LLM and get it to generate a complete immediate, which you’ll then paste into Cursor.
This multi-LLM strategy works effectively once I run into errors whereas vibe coding.
Vibe Coding for Information Science: Yay or Nay?
I personally don’t imagine you can vibe code your approach to production-ready code.
In case you are an information scientist, I nonetheless recommend 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 feel vibe coding has its deserves for particular use circumstances, comparable to constructing an information science portfolio mission if in case you have no front-end coding information.
You too can use it to hurry up your workflows and rapidly study new libraries and strategies you aren’t aware of — 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 may make you a extra environment friendly knowledge scientist:
 
 
Natassha Selvaraj is a self-taught knowledge scientist with a ardour for writing. Natassha writes on the whole lot knowledge science-related, a real grasp of all knowledge matters. You possibly can join together with her on LinkedIn or try her YouTube channel.
