The way I understand it, it's a combination of a unified scheduler for floating point being difficult to implement because many FP instructions need multiple cycles, and FP code being more regular in practice, so you don't need the scheduler to be as powerful.
I'm a mathematics researcher with software development experience, particularly in systems level programming on Linux. For the last five years, I was a math professor (postdoc at UT Austin and Florida State University), but now I want to switch to software engineering. As a mathematician, I'm used to independently solving difficult, often vaguely defined problems, and I really enjoy diving deep into some technical topic and learning lots of new things on the way.
Hence I'm most excited to work on software engineering problems with challenging technical/algorithmic aspects or performance/resource constraints, for example compilers, high performance computing, GPU computing, embedded systems, computer algebra, cryptography, formal verification, symbolic execution, computer security, etc. But I'm open for any opportunity where my skills could be useful!
I'm sure this also depends a lot on the field within mathematics. Areas like mathematical logic or algebra have a pretty formal style anyway, so it's comparatively less effort to translate proofs to lean. I would expect more people from these areas to use proof checkers than say from low-dimensional topology, which has a more "intuitive" style. Of course by "intuitive" I don't mean it's any less rigorous. But a picture proof of some homotopy equivalence is just much harder to translate to something lean can understand than some sequence of inequalities. To add a data point, I'm in geometry/topology and I've never seen or heard of anyone who uses this stuff (so far).
If you are curious, I encourage you to look up The Sphere Eversion Project [1].
It was a project in which we formalised a version of Gromov's (open, ample) h-principle for first order differential relations. Part of the reason it was carried out was to demonstrate what is involved formalising something in differential topology.
Pictures can be directly used in formal proofs. See [1] for instance So, the problem is not in principle, but rather, pictures and more direct syntax has not yet been integrated into current proof systems. When that is done, proofs will be much more natural and the formal proof will be closer to the 'proof in the mind'.
I'm a mathematician with software development experience, particularly in systems level programming on Linux. I have been programming for 20+ years, as a hobby and as part of my work, recently mostly building computer experiments for math research (see my website). I've worked as a professor (postdoc at UT Austin and Florida State University) for the last few years, but I want to switch from academia to industry. As a math researcher, I'm used to solving hard problems without established solutions, and I enjoy learning new things every day.
I'm most interested to work on software engineering with challenging technical/algorithmic aspects or performance/resource constraints, for example: compilers, high performance computing, GPU computing, embedded systems, computer algebra, cryptography, formal verification, symbolic execution, computer security, etc. But I'm open for any opportunity where my skills could be useful!
After I got my PhD in math, I've worked as a university professor (postdoc) for a few years, but now I want to switch from academia to software engineering or related fields. I have 20+ years of coding experience (as part of my job and as a hobby), particularly systems level programming in Linux.
As a math researcher, I'm used to solving hard problems without established solutions, and I enjoy learning new things every day. I'm open for any opportunity where my skills can be useful. Examples I have in mind are working on compilers, high-performance computing, or machine learning (both on the math/statistics side and the algorithms/optimization side), but I'm also happy to hear about anything else!
Hi Florian, my team is hiring ML compiler engineers who are keen to tackle hard problems in the area of scheduling ML graphs (think LLMs, ConvNets, etc.) onto neural processors with limited memory/parallelism. My team's goal is to develop algorithms to minimize either bandwidth or latency for running inference on these graphs. We work on NP hard problems that are approximated by heuristic algorithm approaches and are looking for smart people who can tackle these tough problems.
Location: Tallahassee, FL, USA
Remote: yes
Willing to relocate: no, but can travel occasionally
Technologies: Rust, C/C++, Python, Linux, machine learning
Email: [email protected]
I'm a math professor (postdoc) at Florida State University, and software developer with 20+ years coding experience (as a hobby and as part of my job). I'm looking to transition from academia to a job in software engineering or related fields.
As a mathematician, I'm used to solving hard problems without established solutions, and I enjoy learning new skills. While I'm open to almost anything, I'm particularly excited about work where my math background is useful, like machine learning, data science, quantum computing, or which has an interesting algorithms or performance aspect, like compilers, high-performance computing etc.
I'm a postdoctoral mathematics researcher and professor at Florida State University. I'm looking to leave academia for a job in software engineering, machine learning, data science, or anything where my math skills can be useful.
I have a broad knowledge of math, physics and computer science. I've been programming for most of my life, and regularly use computer experiments in my math work (see examples on my website). These tend to be niche problems, so I often build all algorithms from scratch. I have also worked in the industry a bit before my PhD (building simulation software in C++ and C#).
As a mathematician, I am used to tackling hard problems without established solutions. I enjoy learning new things, and through years of teaching and presenting my research I am also fairly good at communicating complex ideas.
I'm particularly interested in work which has an interesting algorithms or optimization aspect, like high-performance computing, compilers, low-level/embedded programming etc. I also got really interested in machine learning recently, and started reading about and experimenting with it (my math background helps here!). But I'm pretty open-minded, so if you think I could be useful for whatever you're doing, please reach out!
That's just saying git doesn't connect to the internet (or whereever the remote is) to check for updates without you explicitly telling it to. I think that's a desirable property, though the message could be clearer.
I'm a mathematician with 4 years of research and teaching experience post-PhD. I'm looking to leave academia for a job in software engineering, data science, or anything where my math and computer skills can be useful.
As a mathematician, I am used to tackling hard problems without established solutions. I learn quickly, and through years of teaching and presenting my research I am also good at communicating complex ideas. I've been programming for most of my life, and often use computer experiments in my mathematical research. Some projects are on my website. I have also worked in the industry a bit before my PhD (on simulation software in C++ and C#).
I especially enjoy relatively low-level programming, performance optimization, Linux kernel development, embedded devices, and similar things. More recently, I started reading about and experimenting with machine learning (my math background is useful there).
https://chipsandcheese.com/2024/07/15/a-video-interview-with...
The way I understand it, it's a combination of a unified scheduler for floating point being difficult to implement because many FP instructions need multiple cycles, and FP code being more regular in practice, so you don't need the scheduler to be as powerful.