Picture by Creator
# Introduction
Increase your hand in case you’ve ever frozen throughout a technical interview when the interviewer requested, “Stroll me by way of your method.” Most candidates learn the query, leap straight into code, and hope muscle reminiscence kicks in.
What in case you may add an interview query into an AI instrument and get a podcast clarification of your code, a visible thoughts map, flashcards, and a quiz? Effectively, you possibly can with NotebookLM.
On this article, we’ll first remedy Meta’s “Advice System” interview query ourselves, then use NotebookLM’s six options to know and study from it smarter, not more durable.
# Meet NotebookLM: Google’s AI-Powered Research Assistant
NotebookLM is Google’s AI research assistant. It transforms how we study from knowledge by combining a number of AI-powered options that make the method extra interactive and adaptive.
It does so by turning your paperwork, books, or another supplies into interactive studying instruments. Extra particularly, it turns your supplies into conversations, visible maps, and quizzes, and saves hours of guide evaluate, making advanced matters simpler to digest.
This helps you perceive and retain data sooner. Briefly, NotebookLM helps you study smarter, not more durable.
Let’s now remedy the Meta interview query, then discover how NotebookLM may also help us study from that very same code.
# The Problem: Meta’s “Advice System” Interview Query
Advice System
You’re given the listing of Fb mates and the listing of Fb pages that customers comply with. Your process is to create a brand new advice system for Fb. For every Fb person, discover pages that this person would not comply with however at the least considered one of their mates does. Output the person ID and the ID of the web page that must be really useful to this person.
The query is on the market on StrataScratch. On this query, Meta requested us to construct a advice system that implies Fb pages a person doesn’t comply with but, however at the least considered one of their mates does.
// Understanding the Information Behind the Downside
We’ve got two completely different datasets, user_friends and user_pages.
The primary one exhibits every person’s social connections (who’s mates with whom), whereas user_pages lists which pages every person already follows.
Collectively, they assist us discover new web page suggestions primarily based on mutual mates’ exercise.
Let’s preview the user_friends dataset first.
| user_id | friend_id |
|---|---|
| 1 | 2 |
| 1 | 4 |
| 1 | 5 |
| 2 | 1 |
| 2 | 3 |
Let’s preview the second dataset now, user_pages:
| user_id | page_id |
|---|---|
| 1 | 21 |
| 1 | 25 |
| 2 | 25 |
| 2 | 23 |
| 2 | 24 |
// Step-by-Step Answer: Constructing the Advice Engine
The objective is to advocate Fb pages {that a} person hasn’t adopted but, however that at the least considered one of their mates already follows.
First, we join every person with the pages their mates comply with. This helps us see which pages are adopted by the person’s community and potential suggestions.
friends_pages = users_friends.merge(users_pages, left_on='friend_id', right_on='user_id')
friends_pages = friends_pages[['user_id_x', 'page_id']]
Now we have to exclude pages that customers already comply with. We do that by merging once more with the unique users_pages desk and protecting solely the pages that aren’t already within the person’s listing.
comparability = friends_pages.merge(
users_pages,
how='left',
left_on=['user_id_x', 'page_id'],
right_on=['user_id', 'page_id']
)
outcome = comparability[comparison['user_id'].isna()][['user_id_x', 'page_id']]
Lastly, we take away any duplicates to keep away from recommending the identical web page a number of instances and rename the column for readability.
outcome = outcome.drop_duplicates()
outcome = outcome.rename(columns={'user_id_x': 'user_id'})
Listed here are the primary few rows of anticipated output.
| user_id | page_id |
|---|---|
| 1 | 23 |
| 1 | 24 |
| 1 | 28 |
| 3 | 23 |
| 3 | 28 |
Up up to now, we now have relied on conventional methods: coding, merging, and deciphering every part ourselves.
Now, Google’s NotebookLM permits us to take the training course of a step additional.
# From Fixing to Studying: Enter NotebookLM
On this part, we’ll present find out how to present NotebookLM with the main points of the Meta interview query, then discover its six interactive options that show you how to study smarter.
// Step 1: Making a New Pocket book and Including Your Information
Earlier than NotebookLM can help us, we have to present it with details about our interview query. This begins by creating a brand new pocket book. Head to the NotebookLM website and click on “Create a brand new pocket book.”
It is going to ask you for the supply.
As you possibly can see within the screenshot beneath, completely different choices can be found.
- Google Drive
- Web site
- Youtube
- Paste textual content
Let’s use the “Paste textual content” choice and paste the metadata of the Meta’s interview query. Right here is the knowledge we’ll insert utilizing this reusable format.
Right here is the query.
[paste question here]
It makes use of two datasets.
[paste dataset information here]
Right here is the python answer.
[paste python solution here.]
Subsequent, click on on “Insert”, proven beneath.
// Exploring the Interactive Studying Studio
Beneath is the display screen that opens after we hit “Insert.” On the left, we are able to see the sources. Within the center, we now have an LLM that was educated on our knowledge. On the proper, we now have a “Studio” containing six completely different studying codecs, which we’ll check one after the other.
Let’s click on on “How does the Python answer leverage present pal and web page knowledge for suggestions” button, which is underneath the enter textual content within the center. This query was generated by NotebookLM.
Right here is the reply.
As you possibly can see, it explains all the idea. You may as well save this notice by clicking “Save notice” on the finish. The notes will then seem underneath the “Studio” part like this.
So, if in case you have any questions associated to this interview query, you should use the LLM in the course of the earlier display screen to get a solution. Let’s discover the “Studio” six options that make issues extra fascinating.
// Experiencing NotebookLM’s Six Studying Options
On this part, we’ll see NotebookLM’s six studying options, proven beneath, in motion.
Every one helps you perceive the Meta interview query from a distinct angle — by way of audio, video, visuals, and interactive observe.
1. Generate an Audio Overview
Click on “Audio Overview.”
If you do this, you’ll see this notification underneath “Studio,” and your audio overview will likely be prepared in a few minutes.
NotebookLM turns your uploaded content material right into a podcast-style dialog. Two AI voices focus on your drawback such as you’re listening to a tech interview prep present. They break down Meta’s advice system, clarify the logic, and spotlight edge instances.
You’ll be able to obtain this dialog, and apparently, you too can be part of it by clicking on “Interactive.” The display screen beneath will then open.
And if you click on on be part of, the message will say “Howdy, somebody desires to hitch,” permitting you to hitch the dialog. (Sure, you possibly can actually be part of the dialog; it’s up to date in real-time by way of a text-to-speech API. A extremely futuristic function!)
2. Generate the Video Breakdown
Return to the “Studio” panel and hit “Video Overview.” NotebookLM creates a video that visualizes your knowledge and explains the answer. In our case, the video is 6 minutes and 21 seconds lengthy.
Within the generated video, NotebookLM first explains the broader idea: how social media platforms resolve what to advocate, earlier than transferring into our particular drawback.
Subsequent, the generated video discusses the interview drawback by breaking it down into elements, beginning with this key step proven beneath.
Subsequent, it begins explaining the answer. It doesn’t simply repeat what’s on the display screen; the reasons go deep into the idea.
And at last, it breaks down the answer into steps and explains it utilizing these steps.
3. Map the Logic Visually
Flip again to the “Studio” panel and click on the “Thoughts Map.”
NotebookLM generates a visible tree of the issue. Let’s see one.
You see all the drawback construction in a single view: advice objective on the high, required datasets within the center, and answer steps on the backside. Let’s click on on considered one of them, for instance, “Step 1: Determine Buddies’ Pages ( Merge 1).”
As you possibly can see, the thoughts map expands by explaining step 1. You’ll be able to obtain it, too.
4. Construct Studies
Flip again to the Studio panel and click on “Studies.” NotebookLM asks you to pick which sort of report you need to create.
Let’s choose an issue walkthrough. It is going to begin producing a report.
Right here is the step-by-step walkthrough that explains find out how to remedy this interview query.
5. Generate Flashcards
Flip again to the “Studio” panel and click on “Flashcards.” NotebookLM auto-generates Q&A playing cards. Let’s see considered one of them.
And if you click on on the reply, right here is the outcome.
Let’s click on on “Clarify.” It makes use of a immediate to reply this query with Gemini fashions.
Now let’s see the outcomes.
6. Create a Quiz
Flip again to the Studio panel and click on on the Quiz. NotebookLM generates a observe check. Let’s do it and see the primary quiz.
Let’s click on on “Trace.”
The reply is now apparent. So, let’s choose it and see what occurs.
For those who nonetheless have a query, click on on “Clarify.” It takes you to the center part, the place the LLM educated on our answer offers the reply. Right here it’s.
# Conclusion
Prepping for technical interviews does not imply grinding interview issues in isolation. It means understanding deeply, visualizing clearly, and explaining confidently. NotebookLM turns a single interview query right into a multi-sensory studying expertise, together with audio, video, visible maps, written reviews, and lively recall.
You have already got the problem-solving abilities. Now you could have a system to prepare them, reinforce them, and current them underneath stress. Subsequent time you see a tricky SQL or Python query, do not panic; add it, discover it, and grasp it.
Nate Rosidi is a knowledge scientist and in product technique. He is additionally an adjunct professor instructing analytics, and is the founding father of StrataScratch, a platform serving to knowledge scientists put together for his or her interviews with actual interview questions from high firms. Nate writes on the most recent tendencies within the profession market, provides interview recommendation, shares knowledge science initiatives, and covers every part SQL.
