We build AI for physical security and operations.
Daskell is an AI research and engineering team based in Italy. We work with surveillance integrators, operations centers, and critical-infrastructure operators across Europe. Our work is the pipeline that takes a state-of-the-art vision model and lands it, deployed and tuned, on the customer's hardware.
We are a small team — AI researchers, ML engineers, deployment engineers — focused on one problem: making vision systems work, in production, on the hardware our customers actually run.
Most of our work is invisible: model tuning, compile-time optimization, latency budgets, integration with the customer's existing surveillance stack. The visible part — the operator interface, the events that fire — is the smaller half.
The people doing the work.
LorenzoCEO
FedericoCTO
GiovanniML Engineer
AndreaDeployment Engineer
AtabaySoftware Engineer
+ Many AI agentsAlways on
Roster updates with the team. Reach out via info@daskell.com to talk to any of us directly.
What we will and won't ship.
We work in physical security. The systems we build observe people. That responsibility is the first thing on our minds when we accept an engagement. These are the principles we operate by — and the ones we use to decide what we will and won't take on.
On-prem by default.
Customer data does not leave the customer's network. Cloud deployments exist where the customer asks for one, but they are the exception, not the default.
No identity inference without consent.
We detect, track, and classify behavior. We do not perform face-based identity inference unless the deployment is explicitly authorized by the customer, the operator, and the relevant regulators.
Operator in the loop.
Brightsight surfaces events for an operator to act on. It does not take physical actions. The operator decides what happens next; the system supports the decision.
Auditable.
Every detection, track, and event is timestamped, attributed to a model version, and stored in a way the customer can audit. We will not ship a system the customer cannot inspect.
We hire when we have a real seat.
We don't keep an evergreen careers page. If you're a vision researcher, an ML engineer who likes embedded targets, or a deployment engineer who has wrestled an on-prem stack into production — write to us at careers@daskell.com.