You and the process

With a person
Anna, in-house recruiter
9 years hiring engineers
30 minutes with a real recruiter
They read your CV with you, on a call, and say where the offers are being lost.
Didn’t find what you were looking for? Tell us what to build
Datadog

Senior AI Research Engineer - GPU Systems & Kernels

  • Hybrid
  • 6+ years

Salary

Not stated

Similar roles pay €65K - 120K a year · our estimate

AI summary

For members

The whole posting in a few lines. Sign up to read it here and on every role you open.

Sign Up to Read

Description

You will join the Frontier Foundations team as a Research Engineer and work with Research Scientists to make large-scale model training and inference faster and more efficient, close to the limits of the hardware.

You will work at the lowest layer of the stack, writing and tuning GPU kernels and removing memory and communication bottlenecks that slow down large Mixture-of-Experts models on the newest GPUs.

Frontier Foundations builds and runs the high-performance infrastructure used to train and serve large language models at scale, and supports two research areas. The first is Trained Agents for Observability, which post-trains models to act autonomously across the observability domain. SRE incident response is the first target, with planned extension to code repair, security response, and infrastructure optimization. The team builds the simulation environments, RL training loops, and evaluation infrastructure for these agents, aiming to match or beat frontier models at lower cost.

The company AI Research (DAIR) works on high-risk, high-reward problems tied to cloud observability and security. Its earlier AI work includes Bits AI, Bits Evolve, and the Toto time series foundation model.

Responsibilities

  • Write, tune and debug GPU kernels in CUDA, Triton or similar for the operations that dominate LLM training and inference.
  • Optimize code for the latest NVIDIA GPU generations, covering the memory hierarchy, tensor cores, and low-precision formats such as FP8 and FP4.
  • Cut communication cost in distributed workloads such as all-to-all and all-reduce, overlapping it with compute over NVLink and across nodes.
  • Profile end-to-end training and rollout runs to determine whether they are limited by compute, memory or communication, and measure how much room for improvement remains.
  • Build benchmarks, profiling tools and regression tests that inform optimization decisions.
  • Work with Research Scientists to turn new model architectures into efficient implementations.
  • Investigate emerging techniques such as low-precision training and serving, speculative decoding, and fused or persistent kernels.

Requirements

  • You have deep experience in at least one of: GPU kernel development and optimization (CUDA or similar), low-latency communication and collective operations (such as RDMA or NCCL), or large-scale training infrastructure.
  • You can reason about hardware limits (memory, bandwidth, interconnect topology), identify where a bottleneck sits, and explain how you would verify it.
  • You are proficient in a systems language such as C++, Rust or CUDA C, and in Python.
  • You have strong debugging and profiling skills, including with tools such as Nsight.
  • You can explain design and performance trade-offs clearly to technical and non-technical audiences.
  • Bonus points for any of the following:
  • You have experience with Mixture-of-Experts models, expert parallelism, or sparse model inference.
  • You know GPU communication libraries such as NCCL or NVSHMEM, and compute/communication overlap.
  • You have used Triton, CUTLASS or similar kernel frameworks.
  • You are familiar with distributed training frameworks such as Megatron-LM, TorchTitan or DeepSpeed.
  • You have built GPU-intensive systems in graphics, rendering, simulation or scientific computing.
  • You have contributed to open-source ML or systems infrastructure projects.
  • You do not need to meet every qualification on day one. If you enjoy technology and want to build your skills, the company encourages you to apply.

Conditions

Hybrid work model.

Where you’d work

Part of the week in the office

About the company

Datadog

  • Industry: SaaS

Office in Paris, France

Also hiring in New York, United States, Dublin, Ireland, Boston, United States and 8 more places

6 of their 51 open roles are remote

Your chances

Worth a look before you spend an evening tailoring a CV for it.

  • 19 checks run
  • 1 red flag

Still hiring?

13 checks

No red flags

How crowded?

6 checks

1 red flag

Fits Me

How well does this role fit you?

Answer a few questions or drop your CV, and every role gets a fit score with the reasons, this one first.

  • Your field
  • Level
  • Stack
  • Work model
  • Salary floor
  • Must-haves
Details11 facts · Role, Location, Compensation, Company
Tech stack
  • Python
  • C++
  • Rust
  • LLMs
  • CUDA
Seniority
Senior
Industry
SaaS
Specialty
ML
Region
Europe
Pay period
Annual
Show 5 more factsShow less

Role

Category
Data & Analytics
Specialty
ML
Seniority
Senior
Experience
6+ years
Tech stack
  • Python
  • C++
  • Rust
  • LLMs
  • CUDA

Location

Work model
Hybrid
Region
Europe
Office
  • Paris, France

Compensation

Salary
Salary by agreement
Pay period
Annual

Company

Industry
SaaS

Something wrong with this vacancy?

Similar vacancies

Share this vacancy

What's wrong with it?

The employer never sees who reported.

Reason