AI Engineer

Durlston Partners; Greater London; On-site; GBP 120,000 - 180,000; Full time

Benefits offered by this job

Bonus

Flat, non-hierarchical team

Relaxed working environment

Job summary

Durlston Partners LLP is collaborating with a well established proprietary trading firm to build AI agents that operate in live trading environments. You’ll own core parts of the agent platform from infrastructure to safeguards, with impact across the business.

Design and run long-lived agent services on frontier models, craft tool-calling loops, memory and retrieval layers, and ensure they are safe to trust in production through permission controls and robust evaluation.

Qualifications

• 2 to 5 years of software engineering experience, with hands-on LLM/agent projects
• Degree in Computer Science, Software Engineering or related field
• Production-quality Python/TypeScript/Rust on Linux
• Shipped multi-step agentic systems using tool calling
• Strong understanding of LLM security risks and permission boundaries
• Experience with retrieval systems (vector databases, embeddings, RAG) and evaluation tools

Responsibilities

• Design and run long-lived agent services built on frontier models
• Write tool-calling loops, memory and retrieval layers for reliable long sessions
• Ensure safety for production: permissions and prompt-injection safeguards
• Contribute to a flat, flexible team culture and on-call support

Skills

LLM applications

Python

TypeScript

Rust

Linux

Tool-calling

Retrieval systems

Langfuse LangSmith

Security & permissions

Financial markets interest

Education

Bachelor's degree in CS/SE or related field

Tools

Vector databases

Embeddings

RAG

Langfuse

LangSmith

Agent SDKs

Job description

We’re partnering with a well established proprietary trading firm that has been running quantitative and high-frequency strategies for over a decade. The firm trades billions in daily volume across global markets and is now investing seriously in AI, with a mandate to take agentic systems from experimentation into live production.

This is a rare opportunity to build AI agents that do real work inside a trading business, not prototypes that sit on a shelf. You’ll own core parts of the agent platform end to end, from the infrastructure they run on to the safeguards that keep them in check, and see your work used across the firm.

Role overview

You’ll design and run long lived agent services built on frontier models, writing the tool-calling loops, memory and retrieval layers that let them operate reliably over long sessions. A big part of the role is making agents safe to trust in a production setting: permission controls, protection against prompt injection, and an evaluation pipeline that catches regressions before they ship. The culture is flat and informal, with plenty of room to pick up new knowledge across both markets and technology. The team shares an on-call rota to support operations around the clock.

Requirements

• 2 to 5 years of software engineering experience, including hands-on work building LLM applications or agents
• A degree in Computer Science, Software Engineering or a related field
• Production-quality code in Python, TypeScript, Rust or similar, and comfort working in Linux
• Shipped multi-step agentic systems using tool calling, beyond single-prompt applications
• A solid grasp of LLM security risks and how to design permission boundaries around them
• Experience with retrieval systems (vector databases, embeddings, RAG) and evaluation or tracing tools such as Langfuse or LangSmith
• A genuine interest in financial markets
• Bonus: MCP, agent SDKs, graph-based retrieval or high-throughput system design

What they offer

• Highly competitive salary and bonus
• Based in London or APAC (HK or SG)
• A flat, non-hierarchical team with a relaxed working environment
• Direct impact on how AI is used across a growing trading business