DEV Community: Aditya

What OpenAI, Anthropic, Mistral and Meta actually ask in AI engineer interviews

Aditya — Sun, 25 Oct 2026 04:50:35 +0000

Prepping for an AI engineering interview usually looks like this: twelve tabs of Glassdoor, a Reddit thread from 2023, three YouTube videos, a half-read paper, and no idea what this specific company will actually ask.

So I went through 270 public interview reports, the write-ups and videos where candidates name the company and the question, and sorted them by company. The first thing that jumps out: every top lab has its own personality. Prepare for "AI interviews" in general and you'll be ready for none of them.

Here's what each one asks, with the real questions.

OpenAI: design it for real constraints

9 reported questions · 6 senior, 3 mid

OpenAI's list is mostly system design with hard constraints attached: latency, cost, permissions, tenants.

"Design a RAG system that has to meet strict latency and cost targets while serving many tenants."
"Design an agent for searching internal documents that respects permissions, cites sources and abstains when it should." (also reported at Google and Lyft)
"Design a harness for running a tool-using agent safely, and an evaluation platform on top of it."
"Here's a small transformer with a few bugs. Fix them, then add a KV cache for decoding."

What they're testing: can you build something that holds up in production, not just in a demo? "Abstains when it should" and "evaluation platform on top" are the giveaways: they want guardrails and measurement designed in from the start.

Anthropic: do the maths, out loud

10 reported questions · 7 senior, 3 mid

Anthropic's list reads like an infrastructure exam. Expect back-of-envelope estimates on the spot.

"You need to copy 500 GB of weights to 1,000 GPU workers over shared 10 Gbps links. What's the lower bound on time, and what would you build?"
"Estimate an FFN layer's FLOPs, parameter and activation bytes, and the communication from sharding it. Which limit binds: compute, HBM or network?"
"Debug a GRPO training loop. What is the probability ratio, and what goes wrong with it?"
"Turn an LLM into a binary classifier using its log-probs and log-sum-exp."

Here's what a strong answer to the first one looks like:

Each worker has to receive 500 GB through its own 10 Gbps link, so 400 seconds is the floor. Serve everyone from one machine and it takes about 4.6 days. The trick is to make every receiver also upload: chunk the weights and pipeline them down trees, or swarm them BitTorrent-style, and you get close to 400 s. Then cover topology (one copy per rack uplink), checksums and resuming after failures.

What they're testing: whether you think in bytes, bandwidth and bottlenecks before you think in boxes and arrows.

Mistral AI: build it from scratch

16 reported questions, the most of any lab · 10 mid, 5 senior, 1 junior

Mistral goes deep on model internals. You will write code.

"Given logits from a toy language model, implement greedy, top-k and top-p decoding in NumPy."
"Implement batched multi-head self-attention with a causal mask."
"How does RMSNorm differ from LayerNorm, and why do recent models use it?"
"Explain FSDP and ZeRO stages 1, 2 and 3. How do you overlap computation with communication?"
"Your RAG chatbot keeps giving wrong answers. How do you find out why, and fix it?"

What they're testing: whether you know what's inside the model, down to the norms, the KV cache and the optimiser. If you've only ever called an API, this is where it shows.

Meta: agents, end to end

9 reported questions · 7 senior, 2 mid

Meta's senior questions are big, open design problems, mostly about agents and coding assistants.

"Design a code-review agent that combines LLM orchestration with deterministic tools."
"How do you keep a coding agent from leaking private code, both through training and at inference time?"
"Design an LLM coding assistant end to end: data, model and evaluation."
"Implement self-attention, give its time and memory complexity, and describe the efficient variants."

What they're testing: whether you can scope a large system, know where the LLM ends and the deterministic code begins, and treat security as part of the design.

Google: post-training and safe agents

11 reported questions · 6 senior, 4 mid, 1 junior

"How do you detect and prevent reward hacking during LLM post-training?"
"Design LLM summaries of e-commerce reviews that hold up against prompt injection and don't leak personal data."
"Design a scalable service where an AI agent plans work and calls approved tools without ever bypassing deterministic controls."

The one question to prepare for everywhere

"Implement batched multi-head self-attention with a causal mask."

It was reported at Mistral AI, ByteDance, Meta, Apple and Uber, more companies than any other question in the set. Add the KV-cache variants that turn up at OpenAI, Netflix and Mistral, and writing attention from a blank file is the highest-return hour of prep you can do.

How a prep session works on LLM Dojo

I put all of this into LLM Dojo, so preparing for a specific company is one tab instead of twenty. Here's what a session looks like:

1. Open your company's page. Every question candidates reported there, split by level, with a link to each original report.

2. Read the question and answer it out loud. One or two minutes, like the real thing. Out loud is where the gaps show.

3. Open the model answer and compare. Every answer is short enough to say in an interview and lists its sources: papers, books, measurements.

4. Didn't click? Open the lesson. Every topic links to a lesson with interactive labs, so you can see the idea working instead of memorising it: race prefill against decode, size a KV cache with sliders, step through BPE, run GPT-2 token by token in your browser.

5. Let it come back. Sign in (free) and the checks you miss return after 1 day, then 3, then 7, until they stick.

Everything in one place

492 interview questions, junior to senior, each with a model answer
262 of them tagged to the company where candidates reported them, across 58 companies: OpenAI, Anthropic, Mistral, Meta, Google, Microsoft, Amazon, Apple, NVIDIA, Cohere, Scale AI, Databricks, Perplexity and more
7 modules, 41 lessons covering everything those questions test: model internals, inference, prompting, evals, RAG, agents and production
70+ interactive labs and a neural network playground
Spaced review, so what you learn this week is still there on interview day
Free. The interview hub needs no account

👉 llmdojo.dev/interview

Got an interview coming up at a company that isn't covered, or were you asked something that isn't in the bank? Drop it in the comments and I'll add it, with a model answer.

From Clueless Freshman to Tech Intern: My Unexpected Journey

Aditya — Wed, 12 Feb 2025 14:28:08 +0000

Hey there! I'm Aditya, a second-year student at Manipal University Jaipur, and boy, do I have a story for you. Picture this: it's 2023, and I'm fresh out of high school, clutching my JEE Mains results. Good news? I scored pretty well. Bad news? Not quite enough for CS at an NIT. So, there I was, packing my bags for Manipal, a mix of excitement and "what the heck am I doing?" running through my head.

Freshman Year: Dreams vs. Reality

Now, let me tell you, freshman year was a wild ride. I had these grand visions of:

Conquering GSOC
Snagging a fancy internship right off the bat

Reality check: I spent most of that year doing what every other freshman does – figuring out how to adult, making new friends, and let's be honest, a whole lot of "timepass." You know, the usual college shenanigans.

Meanwhile, some of my friends were hitting it out of the park:

One buddy landed an internship at GitLab with a dreamy stipend that made my eyes pop
Another friend cracked GSOC, living the dream I had for myself

Talk about a wake-up call! I was happy for them, sure, but it also lit a fire under me. If they could do it, why couldn't I?

The Summer That Changed Everything

Fast forward to the end of my second semester. Summer break hits – three whole months of freedom. That's when it hit me: I needed to get my act together. So, I did what any confused CS student would do – I dove headfirst into the alphabet soup of tech:

ML
DL
GenAI

You name it, I was trying to learn it.

The Plot Twist

But here's where it gets interesting. I'm chatting with some seniors, all proud of my newfound ML skills, and they drop this bomb:

"ML is great, but only if you're gunning for research or an MS."

Talk about a curveball! There I was, thinking I had it all figured out, and suddenly I'm back at square one.

The Pivot That Paid Off

So, what's a confused sophomore to do? Pivot, of course! I decided to give development a shot. Little did I know, this split-second decision would lead me to juggle multiple internship offers just a few months later – and get this, without even breaking a sweat in interviews!

Sounds too good to be true, right? Trust me, I'm still pinching myself. But if you're curious about how I went from a clueless freshman to a tech intern in high demand, stick around. I'm about to spill all the tea – the ups, the downs, and everything in between.

Whether you're:

Feeling lost in the tech maze
Looking for a good laugh at my expense
Searching for some real-talk advice

I've got you covered.

Ready for this rollercoaster ride? Let's dive in!

The Wake-Up Call: When Reality Slapped Me in the Face

Okay, picture this: I'm scrolling through Instagram, probably liking memes or something, when BAM! My phone blows up with messages. Turns out, my buddy just landed an internship at GitLab with a stipend that made my pocket money look like chump change. And if that wasn't enough, another friend casually drops that he cracked GSOC. You know, just another Tuesday for them, apparently.

Now, don't get me wrong. I was hyped for them. Like, genuinely "jumping-up-and-down" excited. These were my boys making it big! But then... it hit me. That sinking feeling in my stomach, like when you realize you forgot to study for an exam, but way worse.

I looked around my messy dorm room, Cheeto dust on my fingers, and thought, "Dude, what the heck have you been doing?"

Sure, I had a decent 8.9 CGPA. Mom and Dad were proud of that. But let's be real - in the tech world, that's like bringing a spoon to a knife fight. I'd spent more time perfecting my video game skills than my coding ones.

That's when the panic set in. If I didn't get my act together NOW, I'd be watching all my friends land awesome jobs while I... what? Became a professional couch potato?

The thought of disappointing my parents hit me like a truck. They've always believed in me, even after I bombed JEE. (Yeah, that still stings.) I couldn't let this be another "Oh, Aditya tried his best" moment. No way.

I needed a win. A big one. Something to make up for the JEE mess and show my parents that their faith in me wasn't misplaced.

So there I was, staring at my reflection in the black mirror of my laptop screen, a half-eaten pizza beside me. It was time to get real. No more "I'll start coding tomorrow" lies. No more excuses.

This was my wake-up call, and it was screaming louder than my alarm ever could. It was time to answer it, or risk sleeping through the biggest opportunities of my life.

Game on, world. This lazy freshman is about to level up.

My 6-Month Glow-Up: From Broke Couch Potato to Accidental Tech Prodigy

Alright, gather 'round for the most ridiculous tech Cinderella story you've ever heard. Spoiler alert: There's no fairy godmother, just a pirated course and a whole lot of caffeine.

The "I'm Too Broke for This" Moment

So there I was, broke as a joke, when my buddy drops this bomb: "Yo, Harkirat's cohort2 is fire!" I'm like, "Cool story, bro. You gonna lend me the cash for it?" Spoiler: He didn't. But here's where it gets juicy – I, uh, found the course. Yeah, you know what I mean. 🏴‍☠️ (Don't judge me, future me can donate to charity or something.)

JavaScript: The Illegal Edition

Armed with my totally legit (cough) course access, I dove into HTML, CSS, and JavaScript like it was stolen treasure. Which, well... you know. But let me tell you, pirated or not, Harkirat's explanations were so clear, I started dreaming in code. Literally. It was weird.

Backend Shenanigans and TypeScript Tomfoolery

Next up: Node.js and Express. At first, it was like trying to read hieroglyphics while drunk. But those assignments? Pure black magic. Before I knew it, I was slinging TypeScript and Next.js like a caffeinated monkey on a typewriter. Shakespeare, who?

The "Wait, I'm Actually Good at This?" Epiphany

Around week 10, I hear Harkirat say, "You guys are ready for internships!" I almost fell off my stolen... I mean, borrowed chair. Me? The guy who couldn't afford the course? Ready for a real job? Mind. Blown.

Project Chaos: The AI Edition

I went project crazy. Every idea got AI-infused, LLM-powered, and future-proofed. My GitHub looked like Skynet's baby photos.

Hackathon Hero (The Plot Thickens)

Guess who snagged 2nd place at a hackathon? This imposter! 🏆 The judge (some bigshot CTO) was so impressed, he offered me TWO internships. I was half expecting him to ask for my receipt for Harkirat's course. Dodged a bullet there!

Offer Avalanche

Suddenly, I'm drowning in offers. That Peerlist chat? Turned into a job offer. My current startup internship? Old news. Hyderabad startups are throwing money and equity at me like I'm the last developer on Earth.

The ML Resume Mixup

Oh, and because the universe loves a good joke, my hastily made ML resume (yeah, I have two, like some kind of resume-hoarding goblin) caught IIT Bombay's eye. Summer research internship? Sure, why not add "accidental academic" to my fraud... I mean, portfolio.

The Grand Finale (or Beginning of My Villain Origin Story?)

Six months ago, I was a broke kid with more dreams than dollars. Now? I'm juggling offers like a circus performer on Red Bull. From full-stack sorcery to AI wizardry, from "Who's that?" to "We need that guy!" – all thanks to a course I didn't actually pay for. (I'm sorry, Harkirat! I'll make it up to you, I swear!)

So here I am, living proof that with determination, a questionable moral compass, and an internet connection, anything's possible. The view from up here? It's amazing, but I'm a little dizzy.

Moral of the story? Sometimes, you gotta fake it 'til you make it... and then maybe lawyer up, just in case. Now, if you'll excuse me, I need to go figure out how to actually do all these jobs I've accidentally qualified for. Wish me luck! 🚀🤞

P.S. Kids, don't pirate courses. But if you do, at least become a tech prodigy so you can pay it back later. With interest. And maybe a small tropical island as an apology.

Get Started with Machine Learning

Aditya — Fri, 01 Nov 2024 07:45:05 +0000

Getting Started with Machine Learning: A Complete Beginner's Guide 🚀

Have you heard terms like Machine Learning (ML), Artificial Intelligence (AI), and Deep Learning being thrown around everywhere? Don't worry - I'm here to break it all down for you in simple terms!

What This Guide Will Cover 📚

Choosing the right programming language
Dealing with the math (spoiler: it's not as scary as you think!)
A step-by-step learning path for beginners
Real-world machine learning project walkthrough

Let's Talk Programming Languages First 💻

Quick Tip: Choose Python!

Why Python? It's like the Swiss Army knife of machine learning:

🔥 Industry standard
📚 Tons of helpful libraries
🌟 Super beginner-friendly
💼 Most ML jobs require it

The Math Question: Let's Address the Elephant in the Room 🔢

Many beginners worry about the math behind machine learning. Here's the truth: Yes, math is involved, but don't let that scare you!

Important: You don't need to be a math genius to start learning ML. The most important thing is your willingness to learn!

The Machine Learning Project Journey: A Bird's Eye View 🗺️

Let's break down what actually happens in a machine learning project:

Problem Analysis 🤔
- Figure out what you want your model to do
- Could be predictions, recommendations, or content generation
Data Collection 📊
- Gather relevant data
- Could be small or large datasets (sometimes terabytes!)
Data Cleaning 🧹
- Remove inconsistencies
- Fix inaccuracies
- Prepare data for training
Feature Selection 🎯
- Choose relevant information
- Remove unnecessary data
Model Training & Testing ⚙️
- Train your model
- Test its accuracy
- Make improvements

Your Learning Path: Step by Step 📝

1. Master Python Basics First

Learn syntax
Understand objects and classes
Get comfortable with key libraries:
- NumPy: For numerical operations
- Pandas: For data manipulation
- Seaborn: For visualization

2. Start with Simple Algorithms

Follow this order:

Linear Regression 📈
- Perfect first algorithm
- Easy to understand
- Based on simple math
K-Nearest Neighbors (KNN) 🎯
- More advanced
- Great for classification
- Widely used in real applications
Support Vector Machines (SVM) 🔍
- Handles complex data
- Introduces more advanced concepts
- Great stepping stone to advanced topics

3. Advanced Topics

Once you're comfortable with the basics, you can explore:

Q-learning (Reinforcement Learning)
Neural Networks
Computer Vision

Remember! 💡

Learning machine learning is a journey, not a race. Take your time with the basics, and make sure you understand each concept before moving forward. Everyone starts somewhere, and with dedication and practice, you'll be building amazing ML projects in no time!

Happy Learning! If you found this guide helpful, feel free to share it with other beginners starting their ML journey. 🌟