Research Engineer — AI Agent Harness & Eval

Mat Vin; Greater London; Hybrid; GBP 120,000 - 160,000; Full time

Benefits offered by this job

Health insurance

Pension contributions

Hybrid work model

Team lunches and socials

Pet friendly office

Job summary

Tessl, a London-based startup, is hiring a Member of Technical Staff- Research Engineer for Harness Engineering and Agentic Orchestration. You will help shape how agent harnesses orchestrate model behavior, how we evaluate performance, and how production traces inform improvements.

Join a team that works close to customers and ships impactful experiments. The role blends research with hands-on implementation across AI/ML pipelines, evaluation, and tooling, with a hybrid working model in London.

Qualifications

• 4+ years shipping AI/ML products in a startup or applied industry setting.
• Deep expertise in at least one of the four skill areas.
• Strong product and customer instincts; comfortable with customer calls and real workflows.
• Sharp evaluation judgment; knows benchmarks or can prototype assessment methods.
• Experience building datasets for evaluation or training, including pipelines.

Responsibilities

• Sit in on customer sessions and drive fixes through to shipped improvements.
• Close a slice of the outer loop end to end: signal in, dataset out, eval scored, harness change shipped.
• Own a portion of eval infrastructure: dataset curation, harness config, runner, analysis.
• Prototype a new harness or context config and measure impact on real tasks.
• Analyze agent traces and brief the team on actionable insights.
• Collaborate with product and engineering to ship near-term solutions.

Skills

Agent harness

Agentic eval

Outer-loop thinking

Failure-mode analysis

Customer instincts

Evaluation benchmarks

Education

Master's or PhD in a relevant computational field

Tools

Coding agents

Code-generation systems

RL / Bandits

Job description

Tessl, a London-based startup, is hiring a Member of Technical Staff- Research Engineer for Harness Engineering and Agentic Orchestration. You will help shape how agent harnesses orchestrate model behavior, how we evaluate performance, and how production traces inform improvements.

Join a team that works close to customers and ships impactful experiments. The role blends research with hands-on implementation across AI/ML pipelines, evaluation, and tooling, with a hybrid working model in London.