CUDA developer

Join the team of motivated young people building the world’s fastest database for telco data. Enjoy a startup culture with the backing of Instarea – a global big data company. You have an opportunity to join the team at an exciting early stage. Learn more about the product at tellstory.ai

WHAT WE ARE LOOKING FOR:
Good C++ skills are ESSENTIAL and so is familarity with the syntax

Data type system
- basic and compound types, familiarity with layout in memory
- type casting (C++ style)
- type qualifiers
- storage modifiers
Pointers and pointer arithmetic, smart pointers
lvalue-references
The const system
expressions and statements
- operators (familiarity with operation overloading)
- control statements
- name scoping
class basics
- constructor types
- access modifiers
- RAII (Following the RAII principes is crucial)
- the friend modifier
- methods (calling conventions, optional arguments, …)
- basic OOP principles are welcome (inheritance, encapsulation, abstraction, overloading …)
template basics (usage and writing basic tempalte methods and classes)
stl library basic types and operations
- containers
- iterators
- algorithms
exception handling
c++11 and c++14 standard required
- new features and all facilities provided in the new standard
- writing old styled (pre c++11) code is not welcome in our team :)
Familiarity with the commandline and tools for C++ development
- CMake build system
- Basics in bash are an advantage
– Moving around the filesystem and basic tool usage(grep, find, wc, cat …) is sufficient


CUDA KNOWLEDGE:
- Familiarity with basic CUDA constructions
- kernel calls
- CUDA api functions

Familiarity with the CUDA memory hierarchy
- global, local, shared and register memory
- memory transfer operations
- efficient memory access patterns (coalesced access, bank conflicts…)

Knowledge about threading hierarchy
- CUDA threads, blocks and grid layouts

Basics of parallel algorithms knowledge (parallel reductions, map, scatter, gather, prefix sum indexing …)
- knowledge about the nature of these functions is sufficient, we use library implementations, therefore CUB or Thrust library basics knowledge is a must

Parallel optimization basics
- Thread divergence
- CUDA warp optimizations
- exposing more parallelism

Additional facilities provided by CUDA
- atomic instructions
- special instructions
- synchronization primitives

Skills in advanced CUDA topics are welcome but not a required
- register/stream multirocessor occupacy optimizations
- instruction latency hiding
- Knowledge about CUDA profiling tools – nvprof etc
- Git – to know how to rebase is a must have

- Young but experienced team building a global cutting edge technology
- Cool office in the heart of Bratislava, with chilled atmosphere
- Full support for your ideas from the development team and access to cool tools
- The obvious: a competitive salary
- See our cool benefits package https://instarea.com/career
- If you’re not from Bratislava we have a relocation benefit https://instarea.com/career/remote