Senior ML Engineer, AI Infrastructure & Benchmarking

The Senior Machine Learning Engineer  GenAI is responsible for designing, implementing, and operating large-scale systems and tools for AI model benchmarking, optimization, and validation. Unlike a traditional ML Engineer focused primarily on model training, this role centers on building the infrastructure, automation, and services that enable systematic evaluation and performance tuning of LLMs at scale.

This position combines deep understanding of model serving frameworks, GPU optimization, and benchmarking methodologies with strong software engineering skills to deliver reliable, reproducible, and production-grade evaluation pipelines. The engineer will design and maintain validation-as-a-service platforms that allow internal and external stakeholders to assess models across latency, throughput, accuracy, and cost dimensionsintegrating seamlessly with our AI ecosystem and industry-standard GenAI tooling.

A core aspect of this role is creating a robust, extensible benchmarking and

דרישות:

What you ill bring:
Advanced Python for backend development, data processing, and ML/GenAI pipelines.

Kubernetes (Deployments, Services, Ingress) and Helm for large-scale distributed training and inference workloads.

LLM training, fine-tuning, and optimization (PyTorch, DeepSpeed, HF Transformers, LoRA/PEFT).

GPU optimization expertise: CUDA, mixed precision, tensor/sequence parallelism, memory management, and throughput tuning.

High-performance model serving with vLLM, TGI, LMDeploy, Triton, and API-based serving (OpenAI, Mistral, Anthropic).

Benchmarking and evaluation pipelines: dataset preparation, accuracy/latency/throughput measurement, and costperformance tradeoffs.

Multi-model, multi-engine comparative testing for optimal deployment decisions.

Hugging Face Hub for model/dataset management, including private hosting and pipeline integration.

GenAI development tools: OpenAI SDK, LangChain, LlamaIndex, Cursor, Copilot.

Argo CD & Argo Workflows for reproducible, automated