Remote opportunity at
ElevenlabsResearch Engineer - Inference
ElevenLabs is hiring a Research Engineer for inference, focused on putting its cutting-edge AI models into production and tuning them. The engineer owns the path from…
Career Tools
About This Role
ElevenLabs is hiring a Research Engineer for inference, focused on putting its cutting-edge AI models into production and tuning them. The engineer owns the path from a research checkpoint to the serving infrastructure, and builds the systems that carry research advances into live products used by millions of people. ElevenLabs works at the intersection of AI research and product building. It launched in January 2023 with a human-like AI voice…
Job Description
ElevenLabs is hiring a Research Engineer for inference, focused on putting its cutting-edge AI models into production and tuning them. The engineer owns the path from a research checkpoint to the serving infrastructure, and builds the systems that carry research advances into live products used by millions of people.
ElevenLabs works at the intersection of AI research and product building. It launched in January 2023 with a human-like AI voice model, and now counts millions of individual users and thousands of business customers. Its platforms include ElevenAgents for voice and chat agents, ElevenCreative for speech, music, image and video in more than 70 languages, and ElevenAPI for developers. The company says it has raised $781 million and was last valued at $22 billion.
The work covers inference speed and efficiency across the serving stack, including response time, capacity and cost, using methods like quantization, distillation, KV-cache tuning, batching strategies and custom kernels. The engineer also designs fast serving systems for live streaming workloads where timing matters, and creates tooling so researchers can release new models rapidly and carefully, knowing how they will perform. The posting does not require degrees or certifications. Instead, it asks for evidence of solving hard problems, such as earlier projects, design documents or public code, and ideally experience deploying latency-sensitive models, GPU programming and inference optimization with tools like CUDA, Triton, TensorRT, vLLM or SGLang. The job is remote, can be done from any country, and also lists offices in Warsaw, London, New York and San Francisco.
Responsibilities
- Put state-of-the-art models into production and handle the journey from a research checkpoint to live serving
- Optimize inference performance across the stack, covering latency, throughput and cost
- Use methods like quantization, distillation, KV-cache tuning, batching and custom kernels
- Design quick serving systems for live streaming workloads where timing matters
- Build internal tools so scientists can release models into live use fast and safely
Requirements
- Background putting machine learning models into service, preferably where response times are tight
- Solid engineering ability with GPU code and speeding up inference
- Able to work independently on finding and fixing slowdowns anywhere between model design, kernels and orchestration
- Skill in building the instruments that measure performance
Qualifications
- Experience with CUDA, Triton or TensorRT
- Familiarity with inference-serving stacks like vLLM or SGLang
- Proof of tough problems solved, such as earlier work, design documents or public code
Core Skills
Benefits
- Annual discretionary stipend for professional development
- Annual discretionary stipend to meet colleagues in person
- Annual company offsite in a new location each year
- Monthly co-working stipend for those not near a main hub
Frequently Asked Questions
Answers are based only on the employer’s listing; where it doesn’t say, neither do we.
Is the role remote?
Yes. The posting says the job is remote, can be done from any country, and also lists offices in Warsaw, London, New York and San Francisco.
Is a salary stated?
No. The posting does not state a salary or pay range.
Are degrees or certifications required?
No. The posting says it does not need formal credentials, and asks for proof of strong work like earlier projects, design documents or public code.
Which tools are mentioned?
The posting mentions CUDA, Triton, TensorRT, vLLM and SGLang as examples of GPU programming and serving tools.
Sample Interview Questions
AI-generated questions tailored to this specific role — a preview of the full practice set.