Computer Vision Engineer

MBN Solutions; Greater London; Hybrid; GBP 117,000 - 143,000; Full time

Job summary

MBN Solutions is looking for a Founding Applied AI Engineer to develop multimodal agents that operate in real gameplay environments. You will shape the core agent systems, ensuring they handle complex scenarios and break from traditional scripted automation.

An ideal candidate will have experience with AI, computer vision, and agent development, and will thrive in a dynamic startup environment. The starting salary is up to £130k plus equity, with a hybrid work arrangement.

Qualifications

• Experience building multimodal agents that use images and video.
• Strong background in Computer Vision techniques.
• Ability to navigate messy technical environments.

Responsibilities

• Build multimodal agents that operate within live games.
• Ensure agents can handle real-world inconsistencies.
• Shape the core agent system as part of a founding team.

Skills

Multimodal or vision-based agent development

Computer Vision

QA collaboration

Tools

YOLO

ResNET

EfficientNET

Job description

Founding Applied AI Engineer – Agentic Systems (Games) – upto £130k +Equity – Hybrid London

Built agents that actually act in the real world, not just predict?

Worked with multimodal models, images, video, vision, and turned them into systems that make decisions and recover when things go wrong?

Enjoy solving messy, real problems where reliability matters more than demos?

If that sounds like you, keep reading.

The challenge

This is an early stage startup building autonomous agents that test games the way real players do. Not scripted bots. Not brittle automation. Proper agents that observe gameplay, reason over images and video frames, and take actions across real devices.

The goal is simple to say and hard to execute: make studios trust agents more than manual QA.

You’ll be building multimodal agents that operate inside live games, across mobile and desktop, handling inconsistent UIs, timing issues, network lag, and all the things that break naive automation.

This is about getting agents out of notebooks and into production.

What you’ll be building

• Multimodal agents that reason over images and video, often across multiple frames, and decide what to do next
• Vision driven agents that interact with real games, tapping, swiping, clicking, navigating
• Systems that follow loops, observe, decide, act, recover, not fixed scripts
• Automation that runs reliably across devices, OS versions, screen sizes, and edge cases
• Production grade agent systems that people trust to run unsupervised

Think less “prompting” and more systems thinking.

Why this role is different

• Founding level ownership, you are shaping how the core agent system is built
• Real production constraints, not research theatre
• Agents operating in the wild, not controlled demos
• A product that lives or dies on reliability

If you like seeing your work break, fixing it, and making it stronger, you’ll enjoy this.

About you

You’re likely an Applied AI Engineer, Computer Vision Engineer, or similar, and you:

• Have built multimodal or vision based agents using images and video, not just text
• Background in Computer Vision (YOLO, ResNET, EfficientNET etc)
• Familiar with Vision Agent and VLMs (LLaVA, CLIP, Flamingo etc)
• Have worked in similarly messy technical environments, automation, robotics, device control, autonomy, AV, or complex UI driven systems
• Comfortable working with QA’s and Testing Leads to understand their workflows
• Worked in a AI product start-up environment (ideally 0-1)

Interested?

If you’ve built agents that actually do things and want to take real ownership, get in touch.