Autumn AI
Autumn AI is building long-context search infrastructure for agents. Humans effortlessly understand millions of new tokens every week, in realtime. Even the most advanced models can't handle this scale without brittle RAG pipelines. We're a team of engineers and researchers focused on hybrid architectures (state-space + attention). These models fuse memory, search, and reasoning to give your applications human-scale understanding.
To be announced soon
We are planning an open-source release for our 8B-parameter model with its performance on long-context benchmarks.
To be announced soon
We've parnered with a leading healthcare voice AI company to analyze gigabytes of insurance formularies, prior authorization protocols, and clinical guidelines.
We're backed by Y Combinator
Autumn AI is joining YC W26 to scale long-context infrastructure for startups and enterprises building production agents in healthcare, legal, and financial services.
Our focus on long-context engineering
We spent months building retrieval infrastructure for health insurance + prior authorization processing. Our agents kept failing on long-tail queries or when given too much info. Uncovered the same problem while designing end-to-end evals for latency-sensitive voice AI.