AI Engineer (Core Harness)

Granola inc; Greater London; On-site; GBP 70,000 - 110,000; Full time

Benefits offered by this job

Relocation assistance

Job summary

Granola inc is seeking a self-starting engineer to apply the latest LLM advancements to user-centric products. You’ll work on the Core Harness team, turning models into reliable agents and improving how they remember, recover from tool failures, and execute tasks.

You’ll own LLM-based features end-to-end, build evaluation frameworks, and collaborate with a cross-functional team to create a cutting-edge product in London.

Qualifications

• Engineering experience at tech and product-driven companies.

Responsibilities

• Own the development and application of LLM-based features end-to-end.
• Build and maintain evaluation frameworks for prompt iteration and semantic retrieval.
• Optimize embeddings and vector stores for performance and scalability.
• Interact with users to understand their problems and design solutions.
• Stay up-to-date with the latest trends in LLMs and applied AI practices.
• Collaborate with a cross-functional team of engineers, product managers, designers, and other members to create a cutting-edge product.

Skills

LLM solutions

TypeScript

React.js

Node.js

RAG systems

Knowledge graphs

Memory

LLM infra

Tools

OpenAI

Anthropic

Google

Prompt management

Logging/tracing

Eval frameworks

Vector stores

Job description

Hey! We're team Granola

If you haven't already, you should check out what we're building, and why you should work here.

We are looking for a self-starting engineer, passionate about applying the latest advancements in LLMs to create user-centric products.

The Core Harness team builds the systems that turn models into useful, reliable agents at Granola: the agent loop, tools, memory, context management and execution environment. We own how those pieces work together, so agents can understand a user’s work, take action across their tools and follow through on complex tasks.

As an AI Engineer on the team, you’ll work across model behaviour and the systems that shape it. That might mean deciding what an agent should remember, helping it recover when a tool fails, or redesigning its execution loop to make use of a new model’s capabilities.

Why you might be a good fit

• You’ve built with models and investigated their failures. You can explain where a system struggled, how you diagnosed the problem and what made it better.

• You think across the whole agent. You’re interested in how model behaviour, context, memory, tools and execution interact, and can reason about which part needs to change.

• You have judgment about agent/model behaviour. You notice when an agent is being unhelpful, taking unnecessary steps or acting with too little information. You care about making it feel capable, responsive and trustworthy.

• You enjoy a field where the answers keep changing. You try new capabilities, develop your own understanding of their strengths and limitations, and revisit earlier decisions when the evidence changes.

• Love working in a startup environment (you either have experience working in a startup or are really drawn to the zero-to-one phase)

• Are fascinated by LLMs. You love to play with them, figure out what makes them tick and get them to do what you want. You scour the web and reddit for others doing the same (you probably follow @goodside on twitter)

In this role, you will:

• Own the development and application of LLM-based features end-to-end

• Build and maintain evaluation frameworks for prompt iteration and semantic retrieval

• Optimize embeddings and vector stores for performance and scalability

• Interact with users to understand their problems and design solutions

• Stay up-to-date with the latest trends in LLMs and applied AI practices

• Collaborate with a cross-functional team of engineers, product managers, designers, and other members to create a cutting-edge product

Your background looks something like:

• Engineering experience at tech and product-driven companies

• Shipping multi-provider LLM-based solutions to production (using e.g. OpenAI, Anthropic, Google, etc.)

• Proficiency with LLM infra platforms (prompt management, logging/tracing, evals)

• Experience designing large-context LLM systems (RAG, knowledge graphs, hybrid search, memory)

• Building features end-to-end with TypeScript, React.js, and Node.js

What we offer

We are living in the most exciting time for tool builders since Engelbart's demo in 1968. We want to assemble the best crew to build this future together, here in London. Our compensation philosophy is to pay slightly above market on salary and above market on equity.

We do our best work in person, and so our team spends time together five days per week in our new, bright, and spacious office at Old Street. We are happy to offer relocation assistance to candidates who'll be moving to London to join us.

Lastly, we think amazing talent comes from all kinds of life journeys and experiences.