Ai logo

JetBrains AI

Supercharge your tools with AI-powered features inside many JetBrains products

JetBrains AI Hackathon, Powered by Google

At JetBrains, we’re passionate about new technology and always excited to share our latest in-house fun. 

This April, we kicked off our internal AI Hackathon in collaboration with Google and with the support of our internal JetBrains AI and AI Assistant team. It was designed to get our teammates to step out of their daily routines and challenge themselves with ambitious projects. Our partnership with Google gave us exclusive preview access to advanced LLMs like Gemini Pro and Ultra, and we then took things to the next level by using these models in our own projects.

Representatives from Google participated in our hackathon at the Munich and Amsterdam offices. They provided on-site support, served on the jury board, and delivered pre-hackathon enablement sessions and keynotes.

Dmitriy Novakovskiy – Head of Customer Engineering, Google Cloud | Digital Natives, Central & Eastern Europe

It is the most humbling and rewarding feeling for us at Google Cloud to see the great things developers make happen with our technology. With the recent Gemini x JetBrains hackathon, we were happy to see the community of brilliant engineers (at JetBrains) putting our latest Gemini Ultra and Pro 1.5 models through a rigorous test – in the most creative context possible! We were truly inspired by all of the useful projects that came out of just two short days, many of which were truly innovative and “production-worthy”.

The team and I really enjoyed co-hosting this event with the JetBrains team, and we are very much looking forward to doing more of those, always with the “best of Google” technology brought to the table.

In this blog, we’ll share more about this event, reveal the winners, and take a closer look at their amazing projects. Prepare to be inspired!

What, why, and how 


One week before the hackathon, the Google team presented several “pre-hackathon” sessions focused on best practices for using Gemini models. These presentations were designed to get us fired up and ready for the event. 


Once we were all ready to go, the rules and categories for the experiments were announced. 

This hackathon was exclusively for JetBrains employees, and teaming up was allowed. Before the event started, teams could only brainstorm and prepare, no coding was permitted! Then for three days straight, our talented and creative teammates worked on turning their ideas into reality to compete in one of the following categories:

  • User Experience.

For those who wanted to use AI to make developer tools more interactive and personalized. This included things like voice interaction, new interactive elements to enhance daily tasks like tips and activity summaries, and IDE personalization.

  • New features for AI Assistant/ YouTrack with Google Gemini LLMs. 

For those who wanted to design or improve software development tools with Google models. Possible activities in this category included code reviews for projects, new features for the AI Assistant plugin in IntelliJ, ReSharper, and Fleet, and functionality enhancements for YouTrack.

After the Hackathon, the jury chose three winners, who received monetary prizes and some Google-sponsored gifts. Furthermore, every JetBrains employee got to cast their own vote for the winner of the AI Hackathon Trophy.

Winners and their projects

Now for the fun part! In this section, we’ll announce the winners and show you their projects. We’ve also gathered some thoughts on the event from participants, which we’ll share below. 

The AI Hackathon Trophy 

  1. WorkOut


  1. AI Review Assistant
  2. Hey, Fleet! Controlling your IDE through chat/voice
  3. Debug Simulator

Google Models

  1. AI Review Assistant
  2. AI Assignee
  3. WorkOut 

Let’s see what the projects were about! 


Oxana Shumilova, Nikita Volkov, and Alexander Khokhlyavin

Ever feel out of the loop after taking a few days off or when you’re new to a project? Keeping up with the ongoing conversations and updates can be challenging.

That’s why we’ve developed a Slack bot to help you to avoid frustration! With just a simple command like /sum, you can get a customized summary of any channel, thread, or selected discussion group. 

Key benefits:

  • Quick updates: Get briefed on all the essential updates without going through each conversation.
  • Seamless onboarding: New to the team? Get a comprehensive summary to help you find your feet faster.
  • Customization: Choose the channels, threads, and time periods you’re interested in.

Alexander Khokhlyavin

The idea for the project came to me while I was on vacation. So, imagine this: I’m chilling on the beach, minding my own business, when I overhear some work chatter on the phone. Suddenly, all thoughts of sun and sand are replaced by dread about catching up on work stuff when I get back. But instead of letting it ruin my vibe, I think, “Hey, wouldn’t it be awesome if there was a tool that could do all this catching up for me?” And just like that, our AI Hackathon project was born!

Nikita Volkov

We started by surveying users with a Google Form, offering early access to the bot. This formed the basis for our advertising campaign. Then, we refined the concept and functionality. Some ideas came up during casual discussions with colleagues in the kitchen. We realized it was crucial to keep the solution simple and user-friendly, given the limited time. During the hackathon, we went from “Can we really pull this off?” to “This is gonna be epic!” It was a rollercoaster of emotions but it was worth it! 

Oxana Shumilova

During all our Hackathons together, we’ve always kept the positive vibes. So, for our presentation, we wanted to bring that same energy: positivity, humor, and fun. That’s when we had a brilliant idea: Let’s kick off with a hilarious sketch about someone coming back from vacation!

As the clock ticked down to the end of the hackathon, we were in a race against time. Sasha sent me our video just minutes before the deadline. I was also in charge of putting the final touches on the video. Easy, right? Not exactly – the video was taking forever to render. We were rushing, nervous, and on edge. The hackathon presentation started at 13:00, and our video was finally ready at 13:03! But hey – we made it! 

AI Review Assistant

Anastasia Shabalina, Ivan Semenov, Nikolay Rykunov, Egor Bogomolov, and Alexander Koshevoy 

Anastasia Shabalina

The idea of integrating AI into the code review process made perfect sense, as this process is crucial but time-consuming for developers. Initially, our VCS team had about ten ideas for the hackathon and joked about selling them, but it was tough to choose the best. After pitching and voting, the AI Review Assistant didn’t make the initial cut.

Feeling we might be overlooking something, I revisited our list and gathered feedback from 10 random colleagues. Many supported the AI Review Assistant and expressed their eagerness to use it immediately. Encouraged by this feedback, we decided to pursue it at the AI Hackathon. With Ivan Semenov, Nikolay Rykunov, Egor Bogomolov, and Alexander Koshevoy, we refined the idea.

Originally, we planned for AI to handle reviews independently, but it became clear that AI was better suited to assist. Thus, from the initial idea of generating ready-made comments, we expanded the concept. Now, AI Assistant includes a to-do guide plan for reviews with code navigation, AI-generated comments in the editor, and the ability to clarify details directly within the comments.

By the end of the first day, we’d built a working prototype and had all had a great time, complete with music and singing.

Ultimately, the AI Review Assistant emerged as our winning project, directly addressing real user needs and exciting the team. We were thrilled to win 1st place in both categories at the hackathon. We were already planning to bring our AI Review Assistant to production, but this victory has given us extra energy and motivation. We hope our users will soon see the AI Review Assistant in their IDEs.

This image has an empty alt attribute; its file name is AD_4nXfx9dLU0NZvsj3-kUuZ2EFjOtDH1wE1AsYr8tqnJ-RCzrSwkwHTCuvaQIOk_GPnLNrTZKQtPdNxypGONWTfliLyngAPlzZt5uEc1RGkt9X4tqRTsLgHw2xkY6dBF_sgVmzPp0JC9g2W4xihH5ZS1EJX3Ns

Hey, Fleet! Controlling your IDE through chat/voice

Valentin Fondaratov, Alina Mishina, Aleksandr Chernokoz, Artem Bobrov, Titouan Bion, and Vladislav Protasov

The team aimed to enable voice control for Fleet. The goal was to develop a natural language search that comprehends user commands and assists them by changing settings, executing actions, or explaining features based on their spoken requests.

Aleksandr Chernokoz

Over the past year, we’ve been working hard to add AI features to Fleet, with the AI Assistant chat serving as a key starting point for using LLM-based features. We saw the potential for voice input and decided to develop our “voice control” project. The “voice” part refers to accepting spoken commands, but “control” goes deeper – it means allowing the LLM model to manage the IDE by triggering actions or changing settings.

Our project is also interesting because it uses a wake-up word to activate the voice assistant and relies on four different models: a local model for the wake-up word, and remote models for speech-to-text, LLM, and text-to-speech. This setup makes it great for a hackathon, as it breaks down into separate tasks that team members can work on at the same time and then combine when ready.

AI Assignee

Maksim Zheltoukhov and Vladimir Zatsepin

The idea behind the project was to create a digital team member that handles development tasks just like a human colleague. AI Assignee was designed to perform straightforward coding assignments with a bit of guidance.

Here’s how it works! Suppose there’s a task in YouTrack with a title such as “Ensure the FooBar REST API endpoint returns a BadRequest error if the ‘foo’ field is missing.” This task is seamlessly assigned to AI Assignee. Mimicking a typical developer’s workflow, AI Assignee takes charge of the task, asks any necessary clarifying questions, creates a merge request, implements the required fixes, and completes the task once the merge request is approved.

Maxim Zheltoukhov

First of all, I’d like to emphasize what an amazing hackathon this was! Having Google there was incredible, as it gave us access to their advanced large language models (LLMs). These new tools really changed the game, opening up tons of new possibilities and taking our AI Hackathon participation to the next level.

As for our project, it started by wondering what AI could do beyond simple text chats. Could AI handle bigger tasks? This question led to a brainstorming session with Vladimir Zatsepin, where we developed some innovative technical solutions. We equipped the AI with developer tools and showed it how to use them. As a result, we created a system that can write code based on the tasks, manage merge requests, and handle comments. This project pushed AI beyond its current limits and shows a future where AI could become a reliable and innovative helper.

Debug simulator

Elizaveta Shashkova, Pavel Karateev, and Aleksandr Slapoguzov

Using a debugger with real applications can be complicated and slow with lots of additional setups. Debug Simulator simplifies this! Debug Simulator is a tool that interprets debugging commands from an IDE and simulates responses, mimicking real runtime behavior without actual code execution. The team has designed the tool so that it can predict outcomes for variable states, method executions, and more.

Elizaveta Shashkova

We were a fortunate team because all three of us were located in Amsterdam. We came to the office, occupied one of the meeting rooms, and spent two wonderful days hacking together, discussing solutions, and drawing on a whiteboard – just like in the good old pre-COVID times!

Our team had a great variety of skills: Alexander had experience in creating features for AI Assistant, Elizaveta had experience with debugger UIs, and Pavel was our expert in prompt engineering. Therefore, the tasks were naturally split amongst us, and we were able to implement them quickly without stepping on each other’s toes.

Although we conducted a lot of experiments with different prompts, there was still a small sense of fear that once the implementation was complete, the LLM might stop providing us with desirable answers or start to hallucinate at some random moment. We were so surprised and delighted when, in the end, our debugger worked! And it worked exceptionally well on simple code snippets!

Taking part in this hackathon and implementing some fun and unusual ideas was a truly enjoyable experience!

Overall, the AI Hackathon was an amazing experience! Here is what Horia Niculescu had to say about it: 

Horia Niculescu – Customer Engineer – Data, AI, ML, Google

I’ll start by saying a huge kudos to JetBrains for being such a fantastic partner to Google in the AI space. This hackathon is yet another chapter in the incredible collaboration and shared vision between our teams.

During the hackathon I was blown away by the talent at JetBrains. All teams displayed a rare combo of business savviness and technical wizardry. It’s evident in the way they hacked together product features and organizational solutions that genuinely make a difference.

Whether it’s the code writing or code reviewing, the productivity hacks have the potential to boost developer productivity tenfold. My favorite project is the catch-up-on-work-after-vacation assistant – brilliant, fun, and crazy useful. Where do I subscribe?

I’m genuinely excited to see how these projects will be rolled out in production and how this partnership continues to unfold. I’m confident that together, we’ll deliver even more value to our customers in the months and years ahead. Let’s keep pushing the boundaries!

43 projects were presented at the AI Hackathon, and we want to thank everyone for their participation! We’re looking forward to an equally successful event next year! And stay tuned for upcoming JetBrains AI Assistant updates – some amazing new features are on the way, including support for new LLMs. 

image description