Example 4.1
The LeetCode for MLE · MLE / AI interview prep
The interview grind, built for ML engineers.
Abstract. A bank of … interview problems across … topics, from probability and statistics to deep learning, LLMs and coding. Answers are checked automatically, code runs against tests in your browser, and a per-topic skill rating suggests what to practice next.
Keywords: probability, statistics, machine learning, coding
1 Problems are written and verified by the team. Problems reported from real interviews are labelled with the company.
$$\mathbb{E}\big[(y-\hat f)^2\big] = \underbrace{\mathrm{Bias}[\hat f]^2}_{\text{underfit}} + \underbrace{\mathrm{Var}[\hat f]}_{\text{overfit}} + \sigma^2$$
1What’s inside
Every problem has a worked solution and, where the answer can be checked, an automatic grader. Filter by topic, difficulty or role in the bank.
Table 1. The bank at a glance.
–problems
–free to solve
–coding problems
Table 2. Problems by topic. Select a row to open that part of the bank.
2By role
Interview loops weigh topics differently. Start from your role and the bank is filtered to what that loop tends to emphasize.
- 2.1ML EngineerCoding, implementing models from scratch, classical ML and the systems that train and serve them.
- 2.2Data ScientistProbability, statistics, A/B testing, SQL and product metrics.
- 2.3Research ScientistLinear algebra, optimization, probability and deep learning, argued from first principles.
- 2.4Applied ScientistClassical and deep ML, models coded from scratch, and solid algorithms.
- 2.5AI EngineerLLMs and GenAI, production coding, and designing systems around models.
Or follow a study track
3How it works
The practice loop has three parts: answer, check, and choose what comes next.
(a) Solve it on the page
Type a number or an expression, pick an option, or work through a multi-part problem one step at a time, then compare with the worked solution.
(b) Python in the browser
Coding problems open in an editor that runs your Python against test cases on the page. No setup, no install.
(c) A skill rating that steers practice
Each problem you solve updates a per-topic rating and your progress, which is used to suggest the next problem at the right difficulty.
4A worked example
One of the free problems in the bank, exactly as it appears on its page.
5Free and Pro
Table 3. What each plan includes.
Free
$0
- … free problems with full solutions
- Solve in the browser, code included
- Skill rating and progress with a free account
AFrequently asked questions
A.1Which roles does it cover?
ML engineers, data scientists, research scientists, applied scientists and AI engineers. Each problem is tagged with the roles it suits, and you can filter the bank by role.
A.2What is free and what needs Pro?
A set of problems is free, with full solutions, grading and the code editor; the count is in Table 1. Pro unlocks every problem in the bank and its solution. You can browse the free set without an account; an account saves your progress.
A.3How does grading work?
Numeric answers accept equivalent forms, so a fraction, a decimal or an expression that evaluates to the right value all count. Multiple-choice and multi-part problems are checked per part. Coding problems run your function against test cases. Open-ended questions, such as design problems, come with a model answer to compare yours against.
A.4Which languages does the code editor support?
Python, for now. It runs in your browser, so there is nothing to install.
A.5Where do the problems come from?
It is a curated bank written and verified by the team. Problems that were reported from real interviews are labelled with the company they were reported at.
The quickest way to see if it fits is to solve one.