Machine Learning Engineer (LLM Fine-Tuning & Inference) - FT - EMEA/APAC

We’re looking for a Machine Learning Engineer to fine-tune open-source LLMs and deploy them as fast, cost-efficient production services. You’ll own the full path from dataset preparation and training through quantisation, serving and performance optimisation., * Fine-tune open-source LLMs and SLMs (e.g., Llama, Qwen, Mistral, Gemma, DeepSeek) using SFT, LoRA/QLoRA and other PEFT techniques

• Build, clean and validate domain-specific instruction datasets
• Configure and run GPU training, including hyperparameters, mixed precision (BF16/FP16), gradient accumulation and checkpointing
• Evaluate fine-tuned models against baselines using measurable criteria
• Quantise and optimise models for production (AWQ/GPTQ, INT8/INT4, KV cache, continuous batching)
• Deploy and serve self-hosted models with vLLM, TGI, Triton or similar
• Benchmark and improve latency, throughput, GPU utilisation and inference cost
• Expose models through production APIs (FastAPI or similar)

PythonPyTorchHuggingfaceLlm inference tuning, * Strong Python and PyTorch

• Hands-on experience fine-tuning LLMs with Hugging Face Transformers, PEFT and TRL (or Axolotl/Unsloth)
• Solid grasp of transformer architecture and training fundamentals
• Experience deploying LLMs on GPU infrastructure with vLLM, TGI, Triton or equivalent
• Practical experience with quantisation and inference optimisation
• Experience with Docker and Linux environments, * RAG, vector databases and AI agent development
• Automated evaluation pipelines and LLM metrics (hallucination, faithfulness)
• MLOps tooling: MLflow, W&B, Kubernetes, CI/CD for ML
• AWS: SageMaker, EKS, Bedrock
• Distributed training (DeepSpeed, multi-GPU), knowledge distillation, CUDA fundamentals
• Experience with A100/H100/L40S GPUs