Self-supervised learning is pointing the way towards truly intelligent software. Simple objectives like reconstructing missing words or fragments of images are being shown to induce in models creativity, associative memory, an understanding of analogy—traits previously unique to human intelligence.
At Metaphor, we're applying self-supervised learning to search on the internet. Search is our interface to information: it determines the information we see, and therefore what we learn and how we think. A search engine with new capabilities—creativity, memory, analogy—will change the way we learn and think.
We're looking for research scientists and research engineers to join our team. We're a small startup, and you'll help to shape the culture of the place. The hours are usually long and we don't take home a huge salary, but we have a lot of compute per capita and have a good shot at defining a future technology. If that sounds up your alley, here's what we're looking for:
- You have several years of graduate level experience (or are an exceptionally strong undergrad)
- You have broad interests across a variety of domains (NLP, RL, vision, theory, safety)
- You have an eye for fundamental, novel architectures that can scale
- You're comfortable implementing your own ideas and creating large-scale datasets
- (Research engineer applicants only) You're a strong systems programmer with an understanding of networks, performance optimization, and distributed computing
Still apply even if you think you don't fit all the criteria—exceptionally good researchers often don't follow conventional paths.
Metaphor
