12 hours ago
Paris, FranceSenior
Responsibilities
- Build shared LLM-serving foundations, including serving-stack design, GPU capacity planning, and deployment automation for open-weight models.
- Own model registry and promotion processes, including versioning, validation, benchmarking, production promotion, and model catalog maintenance.
- Optimize vLLM, Triton Inference Server, and TensorRT-LLM performance through batching, caching, quantization, and parallelism across the GPU fleet.
- Define SLOs, run launch load tests, implement safe rollouts, rollbacks, and autoscaling, and participate in on-call and incident response.
- Build Datadog dashboards, alerting, and cost reporting for latency, GPU utilization, and inference spend.
- Work with Security and Privacy on network isolation, data residency, auditability, and the ongoing evaluation of serving engines and GPU generations.
Requirements
- 5+ years of software or computer engineering experience, including hands-on production experience with at least one LLM-serving engine such as vLLM, SGLang, TensorRT-LLM, TGI, or Triton.
- Proficiency in Python and Terraform, with a strong understanding of transformer inference mechanics including prefill, decode, KV cache, batching, and GPU compute versus memory constraints.
- Strong experience running GPU workloads on Kubernetes or a managed Ray/Anyscale platform, with performance engineering, reproducible benchmarking, and profiler analysis skills.
- Experience owning production reliability, observability, incident response, and on-call responsibilities.
- Fluent English and strong communication, collaboration, pragmatic build-versus-buy judgment, and ability to work in an evolving scope.
- Preferred experience with KServe, llm-d, NVIDIA Dynamo, disaggregated prefill/decode, KV-cache-aware routing, Mixture-of-Experts serving, multimodal models, or LoRA adapters.
- Preferred GPU kernel or deep profiling experience with CUDA, Triton, or Nsight, or contributions to open-source inference projects.
- Preferred experience in regulated environments and with GDPR, HDS, ISO 27001, or SOC 2.
Benefits
- Free comprehensive basic health insurance for the employee and children.
- 25 days of paid vacation plus up to 14 days of RTT per year.
- Free mental health and coaching services through Moka.care.
- Flexibility days allowing work from abroad for up to 10 days per year.
- Lunch vouchers worth €8.50 per working day, with €4.50 covered by Doctolib.
- Work-council subsidy toward a sports club or creative class membership.
- 50% reimbursement of public transport subscriptions.
- ParentCare birth-leave coverage of 100% salary during the first month and 75% during the second month.
- Enrollment in DoctoGrowth, Doctolib’s long-term employee value-sharing plan.
- Additional support for caregivers and workers with disabilities, including remote-policy adaptations, medical leave days, and psychological support.
- Relocation support for international mobility and access to AI coding tools and dedicated training.
- Permanent, full-time position in Paris with a hybrid setup of up to two remote days per week; start date as soon as possible.
Tech Stack
AWSAzureC++DatadogDockerGoGoogle Cloud PlatformHelmJavaKotlinKubernetesMLflowPythonReact NativeRuby on RailsRustSwiftTerraformTypeScript
Categories
About Doctolib
Doctolib builds a digital healthcare platform for patients and medical professionals, combining online appointment booking, teleconsultations, and practice-management software. It monetizes via SaaS subscriptions to healthcare providers, while access for patients is free. Founded in 2013 and headquartered in Levallois‑Perret, the privately held company operates across France, Germany, Italy, and the Netherlands and is certified as a B Corp.