Robinhood interview is quite focused on communication. Not only do you have to write the correct code, but you also have to explain clearly what you think. The three rounds I interviewed were basically LeetCode medium+system design, and there were no particularly outrageous problems. However, the follow up questions were quite detailed and I had to explain the trade-off clearly.
First round of coding
Topic:Robinhood has a recommendation system: old users recommend new users to join. Each recommendation relationship can only occur once, and there will be no recurring recommendations. We want to know which users have the strongest "recommendation chain", count the total number of new users each user ultimately brings, and then list the top three most influential users.
Problem-solving ideas:First, organize the recommendation data and use a dictionary to record the "recommender-directly recommended user list" to ensure that each new user only belongs to one recommender. Then, for each user, traverse the recommendation chain and count the total number of unique new users. Finally, sort by "descending total number of recommendations, ascending user name", intercept the top three and output them in the format of "user name + number of recommendations". The essence is to deal with the statistical problem of the number of descendants of nodes in "acyclic one-way graphs".
Second round of coding
Topic:Given an order flow consisting of "buy" and "sell" orders, the total number of orders needs to be calculated.
Problem-solving ideas: Use an ordered structure to store unfinished orders: Buy orders are in descending order by price, quickly find high-price buy orders, sell orders are in ascending price order, quickly find low-price sell orders, record "price + remaining quantity", and then process the orders one by one, place buy orders with matching price equal to sell orders, sell orders with matching price ≥ own buy orders, take the smallest amount of transactions and accumulate the total, update or retain the remaining orders, and finally wait for all orders to be processed, and output the total number of transactions.
Round 3: Design a Leaderboard
Topic:Design a leaderboard with the following APIs:addScore(playerId, score) top(K)
Reset(playerId)
This question is actually a data structure design question. I discussed several options with the interviewer at that time:
Option 1:hashmap + sort every time topK. The advantage is simplicity, but the disadvantage is topK's low efficiency under large-scale data.
Option 2:hashmap + minimum heap (size K). When addingScore, the hashmap is updated, and when topK is used, the hashmap is traversed to maintain a min-heap of size K. But there is a pitfall here: if a global heap has been maintained, there may be dirty data in the heap when reset or addScore updates the score. So either lazy deletion or use balanced BST.
Option three:hashmap + TreeMap (or sorted set). Use TreeMap to maintain the mapping of scores to playerId list, so that topK can directly take the largest key. The interviewer agreed with this plan, saying that space was exchanged for time and the thinking was clear.
Final discussion
The interviewer finally asked an open-ended question: What if this leaderboard is used in a high-concurrency scenario? Several directions can be considered:
- Read and write separation
- Cache topK
- Asynchronous batch update
- Distributed rankings (sharding)
This kind of question generally does not require a particularly complete answer, it mainly depends on your system design ideas.
What to do if you get stuck in the interview
In fact, it’s not that many students don’t know how to solve technical questions, but the pressure on the spot is high and their ideas are easily broken. Especially in interviews like Robinhood, where there are a lot of follow-ups, once a certain step is not clear, you may be stuck by questioning all the way.
If you are also worried about this situation happening in VO or technical aspects, you can learn about programhelp in advance. Interview assistance , if you encounter a stuck point during the interview, you can get reminders and direction tips.
