Unum is a startup, so the roles shift a lot.
Skills and experiences are relevant, but so is the ability to learn and jump between subjects. We have an always-open wildcard position for people with solid coding skills in C++/Python and a strong background in mathematics, computer science, or any other STEM subject Please, read the rest of this page and if it feels like you can be part of this - reach us!
What We Do?
|High-Performance Programming||Theoretical Computer Science||Deep Learning & Data-Science|
|Multithreading, Multiprocessing, Async I/O, Linux Kernel Patches, User-Space Drivers, Custom Memory Allocators & Schedulers||Graphs, Hyper-Graphs, Multi-Graphs, B-Trees, LSM-Trees, Skip-Lists, VEB-Trees, Radix-Sort, k-ANN Search, Homomorphic Encryption, Fast Matrix Multiplication, Dynamic/Spectral/Algebraic Graph Theory||Graph Neural Networks, Recurrent Neural Networks, Bayesian Learning, Online Learning, GPT-3, BERT|
How We Do It?
|Python 3||C++ 20||AVX2/512 on x86|
|CUDA 17||NEON/SVE on ARM|
We rarely rely on third-party libraries, but it helps if you are familiar with those:
- Google Test - C++ unit-testing lib,
- Google Benchmark - C++ microbenchmarking lib,
- JAX - XLA-based number-crunching framework,
- PyBind11 - a bridge connecting our Python frontend with C++ backend,
- Conan - Python-based dependency-managemet for C++,
- Docker - containerization software,
- Taskflow - C++ dataflow scheduling across CPU threads and CUDA kernels,
- Halide - LLVM-based C++ library for loop-optimizations,
- PyTorch - a modern general-purpose machine-learning framework,
- Eigen - header-only SIMD-accelerated linear algebra in C++,
- RocksDB - as the only commercially viable LSM implementation on the market,
- ASIO - asynchronous I/O wrappers in C++,
- SPDK or
io_uring- asynchronous I/O avoiding system calls,
- ZeroMQ or NanoMSG - networking without TCP/IP,
- Nvidia Rapids - Python package for GPU-accelerated number-crunching,
- NetworkX - Python package for small-scale network-science,
- BeautifulSoup - Python package for HTML-mining,
- PyMongo - Python interface to MongoDB,
- Flask - Python web-framework,
- Boto3 - Python SaaS integration package.
What We Offer?
- Industry-superior Salaries and Bonuses
- State-of-the-Art Equipment:
- Modern laptops with Gen11 CPUs and RTX 30 Series GPUs
- Densest liquid-cooled servers with 64-core CPUs with Terabytes of RAM, liquid-cooled RTX 30 GPUs and arrays of PCI-E Gen4 SSDs
- Access to the biggest, most interesting and most expensive datasets on the planet, some of which exceed 100 TB in size:
- Time Series
- Flexible Schedule
- Ability to combine PHD Research with work and professional growth in:
- Theoretical Computer Science
- AI & Deep Learning
- Ability to participate in the hiring process and build your dream-team!