This is a hybrid role, based in Sheffield, UK
In this role, you will:
- Design, build, and operate scalable model hosting platforms for LLMs, embeddings, and STT/TTS across heterogeneous hardware
- Optimise inference for latency, throughput, and cost (e.g., quantisation, KV-cache optimisation, dynamic/continuous batching)
- Evaluate and integrate inference frameworks (e.g., vLLM, TensorRT-LLM, SGLang) to maximise performance on target hardware
- Own inference health/performance monitoring (latency, throughput, TTFT, memory, availability) and troubleshoot bottlenecks/deployment issues
- Build end-to-end fine-tuning pipelines (data prep → distributed training → validation) and integrate fine-tuned models into the hosting/inference stack
To be successful in this role you should have the following skills:
- Extensive experience in building AI platforms covering model hosting/inference optimisation and fine-tuning pipelines (LLM experience strongly preferred)
- Strong Python and CUDA engineering; solid understanding of GPU/CPU architecture and HPC fundamentals
- Deep inference optimisation expertise: KV-cache, batching, quantisation (INT4/FP8/GPTQ/AWQ), operator optimisation, framework integration (vLLM/TensorRT-LLM/SGLang)
- Production hosting experience with Docker/Kubernetes and cloud platforms (AWS/GCP/Azure)
- End-to-end fine-tuning expertise: data preparation, distributed training, hyperparameter tuning, HF/Accelerate/LoRA/QLoRA, plus benchmarking/monitoring/troubleshooting
hackajob is partnering with HSBC to fill this position. Create a profile to be automatically considered for this role—and others that match your experience.
Upskill
Level up the hackajob way. Verify your skills, learn brand new ones and test your ability with Pathways, our learning and development platform.
Find out more