Jobs at KOG
All open roles at KOG in one place - salaries, locations, work format and one-click apply.
Jobs from career site
Benefits & perks
Offices & locations
Working at KOG
Kog builds a high-speed LLM inference engine on standard datacenter GPUs, co-designing custom model architectures and low-level GPU kernels for maximum throughput.
What KOG builds
Kog is a real-time AI startup building what it calls the fastest LLM inference engine on standard datacenter GPUs, co-designing the model architecture and execution engine together - its Laneformer model uses Delayed Tensor Parallelism to overlap inter-GPU communication with computation, and its hot path is a handwritten CUDA and HIP monokernel with inline PTX and CDNA assembly. The team hires AI research engineers who reshape open-weight model architectures for inference speed and GPU engineers who write low-level kernels, scaling the stack to large MoE models such as DeepSeek and Qwen.
Frequently asked questions
What does Kog build?
Kog builds a high-speed LLM inference engine for standard datacenter GPUs, co-designing custom model architectures (such as its Laneformer model with Delayed Tensor Parallelism) and handwritten GPU kernels.
Where is Kog based?
Kog is based in Paris, France. Roles are hybrid, with employees spending at least 50% of their time in the Paris office.
Is Kog remote-friendly?
Kog offers a remote-friendly working model, but you will spend at least 50% of your time in its Paris office.
Which roles is Kog hiring for?
AI engineering roles, including an AI Research Engineer focused on LLM inference and architecture research, and a GPU Engineer writing low-level CUDA and HIP kernels.
Does Kog offer equity?
Yes. For GPU engineering roles, compensation is aligned with top technical profiles in the Paris AI market and includes equity.