Gitlab for LLM Teams
Anarchy provides open-source LLM training infrastructure and tooling for developers. We provide a platform to help developers manage their datasets and fine-tunes.
Jobs at Anarchy Labs
Active Founders
Matthew Mirman
Before founding Anarchy, Matt received his PhD in AI from ETH Zurich in 2022 where he scaled verified machine learning from working on 10 neuron toy nets in hours to 100m neuron nets in milliseconds and invented differentiable abstract interpretation and provable training. Before this, he worked at apple on JS compiler optimizations that saved the world a billion dollars in time wasted waiting for pages to load, and before that got an MS from CMU on meta-programming CoC logic languages.
Company Launches
Greetings YC community! We’re Matt and James, the founders of Anarchy. We’ve been working in AI for a long time (Matt just finished a PhD working on formal methods in deep learning) and have become experts in taming LLMs. We started Anarchy to share these advances with devs who want to use LLMs in production.
Problem
While LLMs are beginning to be used as universal text interfaces, developers and prompt engineers are beginning to want consistent and reliable LLMs with agency: LLMs which can search through their ever-changing databases, be able to book meetings, and generally make calls to their APIs. However, coaxing LLMs to behave consistently is black magic requiring constant vigilance, and relying on a single provider can be a risky point of failure.
Solution
Our platform abstracts away the unpredictable work of researching ideal prompts so developers can focus on building business logic: Provide your API and its description, and your LLM-keys, and anarchy will provide you an LLM endpoint (or web-interface chatbot) that can answer questions and solve problems using your API and load-balance between LLMs.
Ask
If this excites you (or anybody you might know, YC or not) and you’d like to try our product, please contact us by signing up.
If you want to have a chat about LLM tech or ask us questions directly, feel free to join our discord: https://discord.gg/qaFf7S373c
We’re also in the process of building up our twitter and linkedin and would appreciate a follow 🙂!
Hear from the founders
What is the core problem you are solving? Why is this a big problem? What made you decide to work on it?
Developers need an easy way to collaborate on the step-by-step process of gathering data and training AI models.Out of the box, LLMs can be >100b parameters and require trillion token datasets. This is more than most organizations are equipped to handle: a few years ago, few had heard of LLMs, and now organizations are scrambling to get ahold of GPUs to train and run them. The famous models are furthermore proprietary. Privacy, quality and efficiency needs are leading organizations to train their own models. \While training these models, developers are constantly adjusting their datasets. As their projects grow, it becomes harder to keep track of the changes, and figure out what data and models to use and track these things in production.
What is your long-term vision? If you truly succeed, what will be different about the world?
We want so solve every problem. We believe mobilizing every developer to work together to building their own AIs is key to this. If you have a unique form of cancer you should be able to spin up your own AI purpose built to solving it, forked from cancer-research-bot.