Roblox's DS summer intern OA has 4 parts: the first two parts are games played on Roblox's own platform, the third part is basically behavioral multiple choice, and the fourth part is coding.
Part 1: the car-building game, 25 minutes total; 2 sections, 13 minutes each, and each section has two different terrains.
You need to equip a car with different parts to get it through both terrains, and there's a limit on how many parts the car can carry.
The goal is to build as many successful cars as possible.
The parts' functions are very abstract — my advice is, if you read the description once and can't understand a part's behavior, don't waste time trying to figure it out from the text. Just slap it on the car and test it, then watch the replay afterward — that's how you actually understand what each part does.
Where you place a part on the car also changes its effect.
There's no great strategy here, because everyone gets different parts and terrains, so you basically just have to keep trying things. Clear the first terrain before worrying about the second.
Installing each part costs energy, and a car can hold at most three parts, so you need to think about how to cover as much function as possible within those 3 parts.
Some parts only work if you use two of them together — I didn't figure this out until the last two minutes, so everyone should watch out for that.
Part 2: the toy factory, 25 minutes total, one section.
It's a factory assembly line making four toys, and you're maximizing revenue over 24 hours.
You buy raw materials, make components, combine components, and finish the toys.
Each toy has a different profit, so you need to work out the cost of the materials/components it needs versus its sale price.
On the line you have to balance each part's production quantity and production time — some parts need other parts to be combined into them, but those other parts have their own different required quantities and production times.
If a part finishes production and overflows storage without being used in time, it just gets wasted.
Each part can only start production once all the required quantities of its input materials have finished producing.
Some parts need a lot of small components, but the max output of those small components is lower than what's needed, so you have to balance a lot of parts against each other.
You can stack a buff on a part's production machine to double its output or reduce its requirement, but that also increases cost.
I personally couldn't come up with any strategy at all — it was just trial and error. But you can try getting one production line fully working first, then move to the next one, starting from the most profitable product.
After finishing this one I honestly felt like my IQ was zero — I never really understood the logic of this game.
Part 3: behavioral multiple choice, 25 minutes, 23 questions.
Each question gives you a scenario and 4 approaches, and you pick the best one and the worst one. It's usually about what you should do when you've spotted an issue or have a different opinion from everyone else. It feels more like a personality test, and I'm honestly not sure what the "correct" answer is supposed to be. But my sense is that the worst option is usually whichever one isn't proactive.
Part 4: coding, 50 minutes, four questions.
It's on CodeSignal, and you can use R or Python.
Question 1: given a numeric array, a significance level, power, and an effect size to detect, use a two-sample z-test to calculate the required sample size.
Question 2: given three arrays — period, group, outcome — calculate difference-in-differences, plus a trend validation check on whether the pre-period difference is within a given threshold.
Question 3: use Bayes' rule to calculate a posterior probability.
Question 4: given a 2D array where each row is a feature, plus an outcome array, run logistic regression and return the names of the top 3 features.
Coding didn't give you the formulas, so you need to memorize the relevant formulas before the test.
Discussion
Loading comments…