Founding AI Engineer

Kantrip; Greater London; On-site; GBP 100,000 - 150,000; Full time

Benefits offered by this job

Office in Bloomsbury

In-office five days a week

Job summary

Kantrip is building an agent system that converts a game spec into a shipped game. You own the system’s next evolution, collaborating with the founder and platform teams to create agents that bridge prompt and playable outputs.

You’ll design, build and operate end-to-end pipelines from spec to release, plus develop evals, retrieval and fine-tuning strategies to optimize quality and cost in production.

Qualifications

• You have built LLM systems in production: multi-step pipelines, tool-using agents, or eval infrastructure.
• You design and run evals: test sets, benchmarks and regression tracking.
• You balance inference cost with quality and latency in a production system.
• You follow the field closely and test new models and techniques as they are released.

Responsibilities

• Design, build and operate the pipeline from game spec to shipped build including planning, code generation, review and release.
• Develop evaluation and regression tooling to measure output quality and catch failures automatically.
• Create the routing layer to select models per step based on quality, latency and cost.
• Replace third-party agent infra with our own runtime where it improves quality or cost.
• Build retrieval and context systems and lead fine-tuning of models on platform-generated data.

Skills

LLM systems in production

Pipeline design

Evaluation infrastructure

Cost management

TypeScript

Tools

Git

CI/CD

API design

Job description

Own the agent system that turns a game spec into a shipped game, along with the evals, the model routing and the runtime underneath it.

The role:

We're building an agent system that turns a game spec into a shipped game. We have a working system but we know it can always be better. You own that system and where it goes next, we can't just use off the shelf system, games need specific loops, context and the economics matter. Model releases lift everyone equally. You'll be building the system that turns those models into games people want to play. You work with the founder and our engine and platform teams to build agents that bridge the gap between prompt and playable.

What you'll do:

• Design, build and operate the pipeline that takes a game from spec to shipped build: planning, code generation, review and release.
• Build evaluation and regression tooling that measures output quality and catches failures automatically.
• Build the routing layer that selects a model for each pipeline step based on quality, latency and cost.
• Replace third-party agent infrastructure with our own runtime where that improves quality, control or cost.
• Build our retrieval and context systems, and lead fine-tuning of models on the data the platform generates.

What we're looking for:

You follow the field closely and test new models and techniques as they're released. You've built LLM systems in production: multi-step pipelines, tool-using agents, or eval infrastructure. Strong systems engineering background. We work in TypeScript; deep experience in another language is fine. You've designed and run evals: test sets, benchmarks, regression tracking. You've managed inference cost alongside quality and latency in a production system. Based in (or willing to relocate to) London. We work in person, five days a week, from our office in Bloomsbury.

Bonus:

• Model routing or gateway infrastructure.
• Fine-tuning, post-training, or retrieval systems.
• A background in games.