We have built ChaiGPT. A conversational AI optimized for entertainment.
You can access our AI via the Chai app. An AI platform - like YouTube but instead of watching videos users speak to AI friends, mentors, imaginary characters.
Chai is the worlds largest conversational-AI platform for entertainment. We serve 500 million AI generated messages a month.
Our mission is building the worlds most entertaining conversational-AI.
Our offices are in Palo Alto, California. You're welcome to drop by and visit if you're in the area.
At Chai, our Research Scientists build great AI products. Applying fancy / latest techniques do not interest us, instead we work to deliver the best user experience (i.e. fine-tuning is not intellectually challenging, but we are happy to do a lot of it if it drives user experience).
The ultimate goal is to deliver the best conversational AI on the planet to our users, making an AI product that our users would like to revisit again and again.
We are not constrained by our tech stack, however we think simplicity is always better. We use a lot of huggingface and PyTorch and everything is in Python.
At Chai, our software engineers need to be very detail-orientated and pragmatic, as we create order out of chaos.
Training and serving large language models is incredibly expensive, user experience can be greatly affected with any downtime or latency in our product. Our software engineers are key for our users having the best product possible, what is the point in advancing these products if they are too expeinsive for our users to have?
We maintain a rapidly developing code base in order to reliably serve our millions of existing users, while ensuring both our front-end and back-end infrastructure is scalable and cost effective. We are not constrained by our tech stack, however we think simplicity is always better. We use a lot of Pytorch, Flutter and Kubernetes.
We think: values > abilities > skills. So primarily we want someone who is a good values fit. This also means we won't be asking you to solve LeetCode problems during the interview. We want someone who is goal driven, hungry for results and can take negative feedback well. This is the only way we can succeed together.
One thing we are not looking for is someone who just wants to write research papers. We care about the product we are building and nothing else.
Achievement. We don't go to work to have fun and play ping pong. We go to work because this will be huge and we want to be the ones to build it.
To succeed as a startup we need to move fast, this is also reflected in our hiring process as we aim to go from application to job offer within 10 days.
The process goes like this: 15 minutes non-technical interview → 45 minutes team interview (walking through your resume and asking you an ML/SWE design question) → one day in-person onsite where you will be given an ML challenge till noon → we all grab lunch together → offer
Simple, deliver great outcomes for the goals we set you together. At a startup, you get no credits for trying, it's all about delivering results.
For software engineers, it could be improving our backend + frontend for our existing millions of users, making it robust and cost effective, delivering the best user experience.
For ML researchers, it could be delivering the best conversational AI on the planet to our users, making an AI product that our users would like to revisit again and again.