AI Research Engineer (London)

Description

The company builds AI agents that learn how a customer's specific processes run, then carry out those operations. Base models are generic, so this is the hard part.

It is a Series A startup backed by Sequoia and other leading VC firms. The co-founders built Forward Deployed AI Engineering at Palantir.

The team is growing in New York and London. It is made up of engineers, AI researchers and strategists who work hands-on and hold a high bar for hiring.

Responsibilities

• Help build a learning system that teaches agents how an enterprise really works. It starts from a customer's existing material, such as knowledge base articles, support chats, tickets, system logs and agent traces, and infers the actual business process from that evidence.
• Turn the inferred process into instructions that a person can review and an agent can run. Include confidence scores so the system knows when to act and when to ask a human.
• Work on the agentic learning environment, where human corrections and agent results feed back into better knowledge and instructions over time.
• Spend most of your time on the core context-learning library and the harness that runs the agents for every customer. The work covers continuous learning, context engineering, agent orchestration and evaluation.
• Contribute to the knowledge benchmarks and to the harness's ability to improve itself.
• Build and extend the core context-learning library, turning problems seen at customers into reusable platform features.
• Design and implement LLM-powered systems and agent workflows, from first idea to production.
• Build agent features that edit, update and maintain large knowledge bases with little supervision.
• Build continuous learning loops that use human corrections and production results to spot knowledge gaps and changed procedures, then improve prompts and instructions.
• Build reliability and confidence systems: evaluation frameworks, confidence scoring, and rules for when to automate and when to hand off to a person.
• Design async, scalable architectures for complex AI orchestration.
• Help shape the company's AI strategy and product capabilities.
• The role mixes applied research with ownership of production systems. Find a failure, form a hypothesis, run the experiment and ship the fix into live customer workflows.

Requirements

• Have built a complex LLM system in production, and be able to walk through the engineering decisions in it, including what changed once real users arrived.
• Have practical depth in at least one of: context engineering, retrieval, agent orchestration, memory, evaluation, or learning from feedback.
• Show sound experimental judgment: turn a vague problem into a testable hypothesis, set a baseline, and separate real gains from noise.
• Enjoy open-ended problems where a solution may not exist yet.
• Move comfortably between research papers and production code.
• Communicate clearly with teammates and customers about your work.
• You do not need to match every point above. The company wants people with different backgrounds and skills, so apply anyway if you are unsure.

Conditions

Compensation is in New York and London. The base salary is competitive.

Health, dental and vision plans are top tier. In the US, the premium is covered 100%. In the UK, private medical and dental cover is provided.

Work in person five days a week from the New York or London office.

The company is an equal opportunity employer and does not discriminate on race, colour, religion, sex, national origin, age, disability, genetic information, marital status, sexual orientation, gender identity, veteran status, or any other legally protected characteristic.

Benefits

• Meaningful equity.
• Flexible paid time off.
• Lunch in the office most days, plus snacks and drinks.
• A stipend to set up your workspace.