Walk into the security office of almost any hotel, apartment tower or shopping centre and you'll find the same thing. A wall of screens. Dozens of tiles. And usually, nobody looking at them. At home it is the same story on a smaller screen: a phone full of motion alerts that nobody opens.
That isn't anyone's fault. No person can watch forty cameras for eight hours and notice the one moment that matters. So cameras became recorders. They capture everything, and they're consulted afterwards, when the damage is done, the report is being written and someone is scrubbing through footage at double speed looking for a face. The cost lands on the owner either way: in time, in worry, or in the one thing nobody noticed.
I started jupyter because I believe that model is finished. The camera is no longer the product. The understanding is. And the understanding should come to you, before you think to ask for it. It is also not reserved for big organisations: the same Helios runs on a single jupyter Hub in a home or a small business, with no subscription for Helios Essentials and no cloud storage fees, because the footage stays on your own Hub.
“AI does the watching. It deters what it can, and it tells you what to do about the rest.”Temi Akinloye
Three ideas
It acts on its own, sees patterns early, and tells you what you did not know.
Everything we build in Helios is measured against three questions. Does it act on its own? Does it see what's coming? Does it tell you something you didn't know?
Unique face clustering
Thousands of sightings become the handful of people who matter.
The most important thing Helios does is also the least visible. Nobody responsible for a place needs a count of detections. They need to know who was there, who came back, and who was new.
Tuesday · 7:00 am · a café before opening
A few hundred people passed the counter yesterday. The owner does not need a number. She needs to know which faces matter.
What jupyter does. Helios counts each person once, shows the regulars and the staff, and brings forward the one unfamiliar face that keeps coming back.
A busy counter, or a busy lobby, produces a flood of detections every day. As a number, that's noise. What anyone responsible for a place actually needs to know is which people matter: how many different people came through, who returns, who is new, and whether any of them are someone you've been asked to look out for.
Helios clusters every face it sees, across every camera and every day, into one identity per person. That single idea unlocks everything else:
- Real footfall. Unique people instead of inflated detection counts, and how today compares with a usual day like it.
- Returning faces. Regular guests, residents, contractors, the cleaner and the neighbour, and, just as clearly, the stranger who keeps coming back.
- Journeys. One person's path from the car park to the lift to level twelve, stitched together from separate cameras.
- Name once. Tell Helios who someone is a single time, and they're recognised on every camera from then on.
It all happens on the Hub on your premises. Faces are grouped where they're seen, not sent off to a distant server to be compared with the world. That is why the footage and the faces stay with whoever is responsible for them, and why there is no cloud storage bill that grows with the number of cameras.
Visual search, watch list, alarm
Describe it, find it, and have the right person told.
Three tools that turn footage into action.
Thursday · 11:50 pm · the loading dock
A contractor’s access card was reported lost this afternoon. Now someone in a hi-vis vest is at the loading dock after hours.
What jupyter does. Describe him and find the clip in seconds. Put his face on the watch list with one box, and the right person hears first if he appears on any camera.
Visual search
Hours become seconds.
"Man in a hi-vis vest at the loading dock." "White van in the basement after midnight." "Someone at the back fence last night." Describe what you remember and Helios finds it, with the matching words highlighted and a picture for every result.
Watch list
Draw one box, and be told if they return.
A banned patron. A known trespasser. The person who tried your side gate. Draw a box around their face in any frame, and you are told if they appear on any camera from that moment on.
Alert and alarm
The right person hears, with the face and the clip.
When someone on the watch list appears, the right person is alerted by push, email or a phone call, with the face, the camera and the clip. With a Halo, the place can respond on its own with a spoken warning or a loud siren. The phone call is part of Helios Pro: a call of ten seconds or less that tells you who, where and the property address. Helios calls you first and, if you do not answer, a number you supply, such as a police station's normal line. It never dials 000, and jupyter is not a monitored alarm service.
The art of the possible
One idea, from a family home to a town centre.
The same Helios runs a family home or a small business on one Hub. Here is what we expect it to become at scale: four kinds of places, one idea. None of it needs a different product, only the same Hub with more doors to mind. Let the AI watch, and let people lead.
Hotels
A lobby that remembers every guest's welcome.
Front of house sees returning guests the moment they arrive, so the welcome feels personal. Security sees the person who has walked every floor's corridor twice tonight without a room key, before a guest complains.
The morning brief tells the duty manager that last night's arrivals ran late because of a delayed flight, the pool gate was left open at 11 pm, and a watch-listed face tried the service entrance.
Apartments and multi-storey accommodation
Hundreds of residents. One calm building.
Residents and their regular visitors fade into the background. Tailgaters at the car park gate, strangers testing doors on the fire stairs and the same unfamiliar van in visitor parking every night rise to the top.
Building managers get patterns, not piles of footage: which entry is busiest and when, which door is propped open and how often, and whether this week is quieter or busier than usual.
Commercial property managers
Every site, understood the same way.
A portfolio manager shouldn't need a control room per building. Each Hub understands its own site, and the insights roll up: which properties are busier than usual, where after-hours activity is rising, and which sites need a patrol tonight.
Contractors are recognised on arrival and departure. Loading docks, plant rooms and rooftops get attention after hours without anyone watching them.
Public safety
Public places, understood by the people who look after them.
Town centres, transport interchanges, parks and council buildings already have cameras. What they lack is time. Helios could give councils, state agencies and police a way to understand these places, not just record them.
Understanding places
See the pattern, and plan from evidence.
How busy a precinct is compared with a usual Friday night. Where people gather after the last train. Which laneway has become a hotspot. Councils can plan lighting, design and services from evidence, not anecdote.
Investigations
From "somewhere on forty cameras" to one clip.
When an incident is reported, visual search and face clustering turn a week of footage review into a sentence and a handful of clips, with a clear record of how each result was found.
People decide
AI recommends. Humans act.
Helios surfaces and suggests. It doesn't dispatch, arrest or judge. It isn't a human-monitored alarm service or an emergency service, and AI can make mistakes, so people stay in charge. It's a tool that gives trained people back the hours they spend watching.
Recommended actions
A brief that tells you what is worth doing next.
The last step is the one we are still building towards. Not just what happened, but what is worth doing about it.
Illustrative examples of where Helios is heading.
Powerful, and responsible
Face clustering, visual search and watch lists are powerful. We take that seriously. Processing happens on premises, on the Hub, so the footage and the faces stay with the family or organisation responsible for them. Watch lists are created deliberately, by people with access, one face at a time. And every alert ends with a person deciding what happens next.
Used well, these tools make places safer and more welcoming at the same time: families and residents are left in peace, guests are greeted, and the rare real threat is noticed early instead of discovered late.
“The future of security isn't more cameras. It's fewer surprises.”Temi Akinloye