Osmo | osmo.ai/careers | Full-time | Machine learning engineer, data analyst, in NYC and other roles
We have digitized vision, and hearing, but not scent — our oldest and deepest sense. Join Osmo on our mission to give computers a sense of smell to improve human health and happiness.
Thrive is passionate about our mission to integrate earlier #cancer detection to empower longer, healthier lives. We are seeking a highly motivated Associate Director/Director, Machine Learning to join our Biostatistics and Informatics team in Cambridge, MA. The successful candidate will be responsible for developing Machine Learning strategies with multi-omics and other types of data, to enable new molecular diagnostic products. The Associate Director/Director will possess deep statistical knowledge, strong analytical skills, solid scientific background, and a high level of professional leadership.
Actually, advanced autodiff is one of its intended points of, er, differentiation :). The authors wrote the original Autograd package [0], released in 2014, that led to “autograd” becoming used as a generic term in PyTorch and other packages. JAX has all of the autodiff operations that Autograd does, including `grad`, `vjp`, `jvp`, etc.
We’re working on the number of supported NumPy ops, which is limited right now, but it’s early days.
Try it out, we’re really excited to see what you build with it!
That’s very cool. Numba and Cython work extremely well with virtually no overhead or extra effort on my part, so jax doesn’t seem like it would buy me much for most of my work. But I can imagine a lot of projects where jax woukd be useful, and I plan to keep current on best practices for it.
var grid = ispace(int2d, { x = 4, y = 4 })
var matrix = region(grid, float)
But beyond that, most things are left to the programmer. E.g., if you want to build a mesh data structure, it's up to you to decide exactly how to build it. (Do you build a graph with pointers? Do you make it an N-dimensional grid? etc.) But my impression talking to application programmers is that most of the time, they want to make those decisions themselves anyway. A standard library might be nice, but it's never going to fit all use cases. So the first job of any language design is to make sure the abstractions are general enough to describe what users want.
@tuananh I've got the dataset stored away, but I don't know if I'm legally allowed to post it. Would love if someone could produce proof one way or the other.
@viraj_shah I spent about 6 months working on the project before I had to stop to concentrate on my schoolwork (I was a senior in collge at the time). I think it would have been impossible to do this for myself without Cython. If it were to happen today, I would probably be writing in PyCuda, or with Numba, and it would be much, much, MUCH more succinct.
We have digitized vision, and hearing, but not scent — our oldest and deepest sense. Join Osmo on our mission to give computers a sense of smell to improve human health and happiness.