How we work with studios
As your outsourcing partner, we take the time to understand your vision and your game before getting into your existing tech stack and requirements. We’re here to answer any questions and to listen to what’s really important to you.
As soon as we have access to your game, we integrate our technology into your engine and focus on designing agents that will learn to play your game. Once we’re done training your AIs and performance is validated, we work with you to make sure your bot system is all set for production.
Your game evolved, it has complete new levels or you have added new game mechanics? Need additional or updated AIs? You’d like bots to replicate tactics your players have developed in the meantime? We can reuse the existing AI and “top up” its training.
Benefits of using ML-based bots
Human-like behaviours mean more engaged players who will spend more time in-game.
We can generate many agents with their own behaviours and skillsets.
Robustness and flexibility to game changes - machine learning-based bots can adapt to their environment without further training which is key for example for procedural maps.
Since our NPCs play your game instead of following a script, they choose their next move based on the player's choices.
We adapt to your own system, whether you are using Unity, Unreal Engine or any other/custom engine, we’ll look for the best way to integrate our technology to your existing system.
You don’t! Not only we’ll work with your team and existing codebase to integrate our bots, but we’ll also make sure you have all the information you need to understand how they fit into your code. Of course, our team will always remain available if needed.
One of the main machine learning technique we’re using is called Reinforcement Learning and relies on the bot learning by playing again and again your game. To get started, all we need is access to your game (so our agents can play it) and a good understanding of your plans and objectives. If you have already some existing gameplay from humans or even simple hardcoded bots this can greatly help the training but is not mandatory.
Training indeed requires a lot of compute and memory but once we have completed this phase and agents are deployed in your game they can be executed on typical video game hardware. We design agents with compute and memory requirements of your target platforms in mind.
That’s up to you and your internal infrastructure: our AIs can either run locally or be hosted on your servers. We will work to design agents that fit the requirements of your platform of choice.
Ideally, we’d be involved as early as possible. To start training our AIs, we need to have access to your game once the main mechanisms and maps are set. Each game is different so be sure to reach out.