ML Systems Engineer: Agentic AI & GraphRAG Benchmarks

Intellectual Capital Resources; City of Edinburgh; Hybrid; GBP 59,000 - 99,000; Full time

Job summary

Intellectual Capital Resources is seeking an experienced AI infrastructure leader for an Edinburgh-based hybrid role, with UK-wide candidates considered. You will own the digital twin software model used to evaluate system behavior before hardware is deployed and will drive agentic AI workloads with measurable system-level improvements.

The role requires strong Python skills, expertise in LLM fine-tuning and RAG, and a track record of designing robust benchmarks.

Qualifications

• Commercial experience in AI systems, with production software shipped.
• Hands-on experience with agentic pipelines, LLM fine-tuning, RAG/GraphRAG, or knowledge graphs.
• Strong Python skills, across ML, distributed systems, and performance engineering.
• Track record of building benchmarks and evaluation frameworks with rigour.
• Systems thinker with high agency and comfort with ambiguity.
• Strong communication skills, able to translate technical results into clear evidence.

Responsibilities

• Own the software model (digital twin) used to evaluate system behaviour ahead of dedicated hardware.
• Build agentic AI and GraphRAG workloads that show measurable system-level improvements.
• Build and maintain a benchmark suite covering latency, GPU utilisation, token reduction, throughput, and cost per query.
• Design experiments that isolate the impact of the semantic memory layer on inference performance.
• Develop enterprise knowledge graph datasets and evaluation methodologies.
• Work with hardware and systems teams to keep software models aligned with hardware capability.
• Generate evidence to support pilots, fundraising, and technical validation.

Skills

AI systems

Agentic pipelines

Python

Benchmarks

Systems thinking

Communication

Job description

Intellectual Capital Resources is seeking an experienced AI infrastructure leader for an Edinburgh-based hybrid role, with UK-wide candidates considered. You will own the digital twin software model used to evaluate system behavior before hardware is deployed and will drive agentic AI workloads with measurable system-level improvements.

The role requires strong Python skills, expertise in LLM fine-tuning and RAG, and a track record of designing robust benchmarks.