Founding CTO

What we do

Flycatcher builds unsupervised visual anomaly detection for industrial inspection. Our system learns what normal looks like from a handful images, then flags everything else in milliseconds, on commodity hardware. No pre-labelled defect data.

The technology comes out of 6+ years of research at IDSIA, a Swiss AI research institute affiliated with USI and SUPSI. We are supported by NVIDIA Inception and funded by Venture Kick, one of Switzerland's main early-stage startup funders, and we are a finalist at Mass Challenge, which TIME and Statista ranked the number two accelerator in the United States in August 2026.

That technology works, and the pipeline is now growing faster than our current team can serve it. What we need is more engineering depth, so we can push harder on delivery instead of choosing which customers to say yes to. That is the turn you would be joining at.

Who you work with

• Dario Mantegazza (CEO, co-founder): owns the research direction and the anomaly detection core. PhD, 6+ years at IDSIA.
• Arianna Marsilio (CPO, co-founder): owns product. Second-time founder, 10+ years in UK startups from early stage to hyper-growth.
• External technical support
  : an ex-Amazon scientist and an EPFL-trained AI engineer, who you will work alongside.

What you own

You are the first senior technical hire, and you are taking over a codebase rather than joining one.

Engineering architecture and delivery. Production systems, from the edge node to the customer dashboard. The training and inference API. Cloud infrastructure and its cost curve. Automation of deployment. Technical hiring. The call on what we build and what we buy.

What you will have achieved

Customers are running our system in paid pilots today. Turning those into production deployments that run every day without us is the work ahead, and leading it is the core of this job. Expect your hands dirty, in the code and in front of customers, for the first few months.

First 90 days

• You have taken the codebase over without a single client delivery slipping.
• You have been on site with customers, delivered our solution to them yourself, and can describe the value we provide in their words rather than ours.
• You have told us where the system will break first, how big that problem is, and what the fix could look like.

By month 6

• You have shipped the API alpha to a partner integrator.
• You have turned a POC into a production deployment that runs unattended and that a customer now relies on. Nobody here has done that before you.
• You have made deployment repeatable and written it down, so the second one costs a fraction of the first, and you have measured configuration hours per facility so we know what we are improving against.
• You have a hiring plan ready to run the moment we grow: roles scoped, bar set, candidates already warm.

By month 12

• You and your team are running five POCs and the first live facilities without a founder in the delivery loop, and you have opened our first manufacturing deployments.
• You oversaw the launch of the Automation Engine alpha working internally, measured against the baseline you set in the first six months.
• Your first three to five engineers hired and productive, with a real on-call and incident process.

By month 24

• You took the Automation Engine and the plug-and-play product in beta.
• You have put API V1 in public hands: authentication, key management, rate limiting, and a support process you built on purpose.
• You have around fifteen facilities running, and manufacturing deployments that no longer need a bespoke rebuild.
• You have driven infrastructure cost per facility low enough that our licence pricing holds its margin at scale.
• You have built an engineering team that hires, ships and recovers from incidents without a founder in the room.

The stack: loose on tools, tight on outcomes

Today we are Python-native; our proprietary models are written using PyTorch, the services are FastAPI, currently we use Azure for cloud: blob storage for models, SQL for customer data, GPU compute for inference.

What sits at the customer end is often not our choice. One site will be a Jetson we install, the next a server in their rack, the next something in their cloud. You will find out on arrival.

That is what we run, not what you must keep. Part of what we are hiring is your judgement about what should change.

FFmpeg, OpenCV and the rest are the barebone of the vision stack.

What we think this needs

The technical bar below is high, but it is not what makes this role hard to fill. Plenty of people clear it. The rarer part is the first three.

• You can carry the company, not only the code. You can sit with a terminal operator, understand what their day costs them, and turn that into a deployment they will pay for. You can represent us to a partner or an investor without a founder translating.
• You build the team, not only the system. You will hire the people who come after you, set the bar they are measured against. You know what a rushed early hire costs, and you…