algobot

AlgoBot 🐍🤖

A Zulip bot made for The Recurse Center to help individuals get better at data structures and algorithms!

AlgoBot has three primary features:

1. Solo Sessions - Setting up a study schedule and providing daily questions through Zulip private message.
2. Mock Interviews - Pairing similarly experienced individuals for mock interviews.
3. Daily Questions - Posting daily questions to Zulip threads for the community to work on together.

Please note that AlgoBot is a direct fork of thwidge/pairing-bot. This project would not have been possible without the hard work put in by Maren Beam and the rest of the contributors over there. AlgoBot says thank you 🙏

Table of Contents

1. Usage
   1. Commands
   2. Solo Sessions
   3. Mock Interviews
   4. Daily Questions
2. Design
   1. Stack
   2. Matching Algorithm
   3. Deployment
3. Contributing
   1. Pull Requests
4. Before Your Mock Interview
   1. What should I prepare for as the interviewer?
   2. What should I prepare for as the interviewee?
5. FAQ
   1. Why do I need to add myself back to the mock interview queue after each session?
   2. I'm a beginner to these types of problems. Can I still pair?
   3. Can I just practice my interviewer skills without being interviewed?
   4. Why should I use AlgoBot over XYZ?
   5. Do I need to code in Python?
6. Additional Resources
   1. Mock Interviews
   2. Additional Questions
   3. DS&A Courses / MOOCS

1. Usage

1.i. Commands

Upon typing help in a private chat with AlgoBot, you'll receive the a response informing you of the following supported commands:

• subscribe to get started! You'll start getting daily data structures and algorithms questions. Oh how fun!
• schedule to add yourself to the queue for a mock interview!
  • You'll remain in the queue until you get a match! Upon interviewing, you'll need to schedule once again.
  • In the case you no longer can mock interview, please cancel.
• skip to skip tomorrow's daily question.
  • unskip if you change your mind.
• config to review and modify your current settings
• unsubscribe to part ways with AlgoBot. Note that your settings and session history will be deleted!

Note that these commands only work in a 1-on-1 chat with AlgoBot.

1.ii. Solo Sessions

The bread and butter of AlgoBot, solo sessions are the questions you receive each day as part of your structured study plan. Questions will be selected from either a problem set or at random based on your configuration. Feel free to treat these as seriously as you'd like; it's entirely up to you whether they act as serious interview prep, a fun exercise to work on with friends, or something in between.

Upon using the subscribe cmd, your account will be assigned the following default configurations:

• Days: Mon/Tue/Wed/Thu/Fri
• Difficulty: Easy / Medium (randomly selected between the two)
• Topics: All / Random
• Problem Set: Top Interview Questions (LeetCode)

These defaults can be viewed and altered at any time using the config option. Note that questions are sent out at 07:00AM EST on the scheduled day so any changes or skip cmds will need to be made before then.

1.iii. Mock Interviews

When you decide that you want to take a stab at actual interviews, use mock interviews to get in that practice with other Recursers. Questions will be selected from either a problem set or by you manually; if you pick your own question, make sure to let your partner know! These sessions are meant to simulate real world interviews; treat them like phone screens and you'll find success!

Upon using the schedule cmd, you will be placed in the queue with other Recursers interested in mock interviewing. You will be assigned the following defaults:

• Days: You're in the queue until you are matched; upon matching, discuss your availability your partner.
• Experience: Medium (you will prepare questions of medium difficulty and below as an interviewer).
• Difficulty: Easy / Medium (randomly selected between the two)
• Topics: All / Random
• Problem Set: Top Interview Questions (LeetCode)
• Environment: LeetCode

These defaults can be viewed and altered at any time using the config option. Note that matches are made and sent out at 05:00AM EST each day so any changes or cancel cmds will need to be made before then.

It is important to note that matches are made based on similarity of profiles to ensure equitable, rewarding interviews. Please take the time to review your config to ensure it matches your preferences and experience level.

If you do not get a match the first day, do not worry as you'll stay in the queue until you do. Upon matching, you'll be removed from the queue. If you want back-to-back interviews, you'll need to manually schedule all over again. Read the FAQ for more details.

1.iv. Daily Questions

For those that want to treat these questions as a collaborative effort, daily questions are posted to a Zulip thread. The difficulty of these questions increases throughout the week (akin to something like the NYT crossword).

Link to thread: Daily Question (updated daily at 09:00AM EST)

Note that you do not require any special configuration or messaging of AlgoBot for these problems. Simply go to the link, try your hand at the question, and post your solution in the thread to discuss with others. Remember to use spoiler tags to prevent ruining the solution for others!

2. Design

2.i. Stack

• Go
  • Deals with communication with Zulip and server.
• Python
  • Scrapes question banks and problem sets.
• HTML/CSS/JavaScript
  • Validates and displays user configuration / session history.
• Google Cloud App Engine
  • Configures serverless deployment and enables versioning.
• Google Cloud Firestore
  • NoSQL Document Database to store question data, user configurations, and session history.
• Google Cloud Cloud Build
  • CI/CD to keep server instance in alignment with master branch.

2.ii. Matching Algorithm

Given that this is a bot meant to improve data structures and algorithms skills, I thought it would be prudent to discuss how matching for mock interviews is implemented.

The algorithm behind the scenes takes inspiration from "The Stable Roommates Problem", a variation on "The Stable Marriage Problem". Instead of using two separate groups and creating a bipartite graph, we consider the population of Recursers as one, uniform group.

Akin to other stable matching algorithms, use preference lists for each individual; interview preference is calculated by how similar user configurations are and how likely an individual is to be able to conduct a productive interview for the user.

Experience and comfort with DS&A come into play here. Someone at an intermediate level prefer someone at the same or higher level to interview them as opposed to a beginner. In the case of preference ties, we just randomly select a winner between the two.

The rest of the algorithm follows the process laid out by Robert W. Irving in his paper, "An Efficient Algorithm for the 'Stable Roommates' Problem".

I'm still trying to improve upon the algorithm moving forward. For one, I'd like to add weighting based on time spent in the queue, ensuring that individuals that have not been matched for some time are given higher priority. Additionally, Irving throws out the results if no stable matching is present. I'd like to still match any stable pairs to ensure atleast a portion of the group is matched.

These modifications are constantly in development and will be introduced to the algorithm over time!

2.iii. Deployment

Just a few details about the setup and deployment of AlgoBot:

• Zulip has bot types. AlgoBot is of type outgoing webhook.
  • This allows us to be notified when certain types of messages are sent in Zulip.
  • Upon a trigger event occurring, a POST request is sent to the server and evaluated.
    • The official documentation does a great job of clarfying all capabilities.
• The database is prepopulated with:
  • An auth token (which the bot uses to validate incoming webhook requests).
  • An API key (which the bot uses to send private messages to Zulip users).
    • Both can be found in your Zulip settings (go to Settings -> Your Bots -> "Copy zuliprc").
• All configuration of App Engine is done through app.yaml.
  • Credentials for Google Cloud are either saved in a hidden JSON or are saved as environment variables.
  • cron.yaml and cloudbuild.yaml configure cronjobs and CI/CD, respectively.

3. Contributing

3.i. What you can do to help

This is meant to be a project built by and for the Recurse Center community so feel free to contribute!

I'll be posting issues and/or feature ideas to the issues board as they arise so take a look there. If you've found a bug or think of a feature that would improve the AlgoBot experience, feel free to add your own issue. Before actually contributing, comment on the issue you're tackling and I'll assign it to you.

In the case you can't find an issue, bug, or feature to work on, here's a list of ideas (listed by language):

Your contributions are welcome and encouraged, no matter your prior experience!

3.ii. Pull Requests

The workflow below has proven to be useful with other projects but please let me know if I can clarify anything!

1. Create an issue
2. Fork the repo
3. Create a branch*
4. Make your changes
5. Write unit tests as applicable
6. Format the codebase using 'go fmt'*
7. Ensure that your changes passes all tests using 'go test'
8. Squash your changes to as few commits as possible*
9. Make a pull request*

*Please use the issue number and name when possible to improve clarity and project maintainability.
Additionally, please adhere to Conventional Commits standards.

4. Before Your Mock Interview

4.i. What should I expect or prepare for as the interviewer?

Upon getting matched, your responsibility as the interviewer is to prepare your assigned question.

Please take the time to really learn the question. The best way to go about this is to tackle the problem yourself!

• Implement the brute force solution and work towards improving it.
• Learn multiple solutions and understand the time and space complexity of each.
• Read the official LeetCode solution (if available), the top solutions in the Discussions section, and/or watch YouTube walkthroughs.
  • I'd recommend Nick White, Kevin Naughton Jr., and Back To Back SWE for good video resources.

If your interviewee seems to be stuck or going in the wrong direction, try to carefully guide them on the right path. As your interviewee implements their solution, ask for clarifications and elaborations on their thought process.

The style of interviewer you are is dependent on what your partner is looking for so when in doubt, just communicate and see exactly what they are looking for out of the experience.

Nobody is expecting you to be a phenomenal interviewer so don't feel overly pressured by this role! By practicing and understanding the question beforehand, you've done your homework and become a valuable resource to another individual tackling the problem for the first time.

Explaining and guiding someone through a problem reinforces your own knowledge so its a win-win!

4.ii. What should I expect or prepare for as the interviewee?

As the interviewee, your only job is to treat the experience seriously. While you'd ideally like to code up the optimal solution, remember that communication is key!

Here are some of the things you should discuss or do as you work through the problem:

• Ask clarifying questions about the problem
  • Even if it's straightforward, take the time to show that you're thinking about all aspects of the problem!
• Walk through an example to reinforce and show your understanding of the task at hand.
• Discuss your initial thoughts; if you see a naive solution, bring it up and identify any bottlenecks.
• Before actually coding, discuss your algorithm at a high level and ensure it makes sense to the interviewer.
• When actually coding, remember to talk out what you're doing!
  • Try your best to use clean variable names and DRY code.
  • Remember that readability always counts more than some flashy trick or syntactic sugar.
• Upon finishing your implementation, walk through it and ensure that it makes sense.
• Run your solution through some test inputs to ensure validity.
• Discuss the space and time complexity using Big O and/or Big Theta notation.
• Review ways you would test your algorithm, ensuring that you've covered all edge cases and branching paths.
  • If you have time, feel free to actually write them out. Otherwise, a high level discussion suffices.

While this list comprises the vast majority of what you should be doing as the interviewee, please discuss any other requirements or goals with your interviewer beforehand! This is your experience so you should have a say as to how it's conducted!

5. FAQ

5.i. Why do I need to add myself back to the mock interview queue after each session?

I decided that it's best that people opt in each time they want to interview to prevent the situation where people forget and possibly ruin the experience for their partner. I think the decision makes sense since interviewing is less common and frequent than something like pairing but I'd be happy to listen to any alternatives y'all might have!

5.ii. I'm a beginner to these types of problems. Can I still pair?

Most definitely! A big part of data structures and algorithms problems is repetition and pattern recognition; how are you going to recognize the patterns without some exposure? The only thing I ask of you is to change your configuration to reflect your current experience level so you can be paired with someone of a similar experience level (using config). I'd also highly recommend taking a course or MOOC about the topic if you have the time (see 5.iv. for recommendations)! In the case you just aren't understanding the question you're supposed to prepare as the interviewer, talk to your partner and see if you can come up with an alternative.

5.iii. Can I just practice my interviewer skills without being interviewed?

Bless your heart! Just set your configuration to match the experience level you wish to interview and let your partner know that they don't need to prepare anything. You sound like a good interviewer :)

5.iv. Why should I use AlgoBot over XYZ?

You should not! AlgoBot is a supplement to the rest of your prep and a great way to get to work with your fellow Recursers in a bit of a different setting. Check out the additional resources I've linked below.

5.v. Do I need to code in Python 🐍 🐍 🐍 ?

I'll be very sad if you don't but disregard my feelings and use whatever language best fits your preferences. In the case your interviewer doesn't know the language you're using, try to explain language-specific features as you use them. I thought of adding a preferred language as a configuration feature but ultimately decided against it as it would further splinter participants and make matching harder.

6. Additional Resources

6.i. Mock Interviews

• Pramp
  • The primary inspiration for this tool, Pramp is completely free and great for getting paired with similarly experienced devs prepping for interviews. I've heard the question pool is a bit limited but it's been solid in my experience.
• Recurse Career Services
  • Your best bet for a realistic interview. Talk to the team and I'm sure they'll get you a great match!

6.ii. Additional Questions (ranked by easiest to hardest)

• Codewars
  • Not entirely focused on DS&A but I've found it to be a great intro to small, challenging problems.
• Firecode.io
  • A very solid resource that uses repetition to ensure you have a grasp on certain concepts before moving forward. The one caveat is that it supports Java/C/C++ and not Python. Hsssss 🐍!
• CodeSignal
  • Somewhat similar to Codewars but it has a great section on common DS&A problems.
• interviews.school
  • Just a curated list of LeetCode problems but it does a good job of listing out what you need to know.
• Advent of Code
  • I love me some AoC. While these problem don't explicitly require efficient solutions, I often code up the brute force solution and then improve upon them. The subreddit has some real coding wizards.
• HackerRank
  • Pretty much on par with LeetCode but does require you to read in input! Some online assessments and phone screens are done here so might be worth checking out!
• LeetCode
  • Hello darkness my old friend. AlgoBot questions are pulled from here so you know it's good.

6.iii. DS&A Courses / MOOCS

• Cracking the Coding Interview
  • Not a course but the standard recommendation (and for good reason). I think its significantly easier than LeetCode but worth checking out. Yer a wizard Gayle Laakmann McDowell!
• Princeton Algorithms
  • I've only done a few psets from this course and have dabbled in parts of the textbook but it's definitely some good stuff. Coursera constantly runs free courses for both parts!
• Berkeley CS61B
  • Josh Hug is one of the greatest professors I've had the pleasure of learning from (also shout out to David Malan). This course has a fair amount of overlap with Sedgewick's course since Prof. Hug taught at Princeton for a bit. Some of the projects, like generating random maps for a Dwarf Fortress-esque game or writing your own Git, were incredibly fun and challenging. The Spring 2018 version released its autograder so you can treat it like a real course!
• MIT 6.006
  • I've only watched lectures from this course but Eric Demain and Srini Devadas are incredible at what they do. Some of the material might be a bit niche for interviews and the lectures can get a bit "mathy" but if you're looking to strengthen your understanding of DS&A, look no further!