Staff Research Engineer: AI Agent Evals & Post-Training

Nominal; Greater London; On-site; GBP 90,000 - 140,000; Full time

Benefits offered by this job

100% coverage of medical, dental, and

Unlimited PTO and sick leave

Free lunch, snacks, and coffee

Professional Development Stipend

Annual company retreat

Job summary

Nominal is seeking a Staff Research Engineer to shape how its agents are evaluated and to build the path from evals to post-trained models for hardware tasks. You will lead evaluation strategies grounded in real-world workloads and coordinate with hardware experts.

You will design benchmarks, drive post-training approaches, and advance ML tooling to operate across air-gapped environments when needed.

Qualifications

• 8+ years in ML engineering or research with evals or post-training leadership.
• Statistical rigor in evals; detect real differences.
• Experience evaluating LLM/agent systems and their limits.
• Hands-on post-training: fine-tuning, RL, distillation on real tasks.

Responsibilities

• Build eval suites for agents, the MCP tool layer, and internal company agent.
• Invent benchmarks for agents using physical engineering data.
• Benchmark our agents against frontier agents using MCP and show advantages.
• Build the eval infrastructure: datasets, model-based and human graders, CI gates.
• Turn evals into reward signals and training data; lead post-training (fine-tuning, RL, distillation).
• Set Nominal's strategy for measurement and model improvement with hardware experts.

Skills

8+ years in ML engineering or research

evals or post-training leadership

statistical rigor in evals

LLM/agent systems evaluation

post-training techniques (fine-tuning,

distillation / RL from feedback

team technical direction

Tools

Claude Code

Cursor

Codex

Job description

Nominal is seeking a Staff Research Engineer to shape how its agents are evaluated and to build the path from evals to post-trained models for hardware tasks. You will lead evaluation strategies grounded in real-world workloads and coordinate with hardware experts.

You will design benchmarks, drive post-training approaches, and advance ML tooling to operate across air-gapped environments when needed.