Every CCTV camera, a trained pair of eyes.

SIMIS is an on-premises AI safety layer for the CCTV a site already owns. It flags falls, collisions, hitting and kicking to staff within seconds, wherever people are in someone else’s care.

No facial recognition. Processed on site. Nothing leaves the building.

1st place, AI Singapore Pan-SEA Developer Challenge 2025 · SUTD ARISE Validate, 2026

Every room watched, one moment flagged, the report drafted before anyone asks.

Your CCTV records everything. You can use almost none of it.

Cameras are everywhere. Awareness, search and a usable record are not. Three gaps; SIMIS closes each one.

One person in a dim office watching a wall of sixteen camera feeds
Sixteen rooms, one screen, nobody watching. That is every recorder today.

Recorded, not seen.

CCTV runs all day. Almost none of it is watched. Incidents happen in the gap between what was filmed and what was noticed.

Covered by SIMIS Monitor

Archived, not searchable.

Finding the moment means scrubbing a timeline by hand. Manual, linear, and far slower than the question deserves.

Covered by SIMIS Search

Forensic by accident.

The write-up is whatever can be reconstructed afterwards from memory and a grainy clip off a hard drive.

Covered by SIMIS Prove

Three layers, from live alert to a record you can show.

Real-time monitoring, searchable history and a record you can hand over. Each layer feeds the next, on the CCTV a site already has.

Monitor

Safety events, surfaced as they happen.

Runs on the CCTV a site already owns. Flags falls, collisions, hitting and kicking to the person on duty; people decide. No facial recognition, nothing leaves the premises.

Search

Find footage by what happened, not when.

Flagged events are indexed, so you ask in plain language. “Two people near the back gate, after 3pm” returns clips in seconds, not a timeline to scrub.

Prove

Written record, ready when required.

Every confirmed event becomes a structured record with the clip attached and the review history intact. When a family, inspector or insurer asks, you export a document.

What operators receive.

Reports, alerts and tamper-evident records: what settles the conversation after an incident. Shown for an early-childhood site; the outputs are the same in every setting.

Built for places with a duty of care.

Wherever one person is responsible for another, the same moments matter: a fall nobody saw, a collision, a hit or a kick. One engine, configured per setting.

Early childhood

Preschools, infant care, student care

  • A fall on the mat
  • A collision between children
  • Hitting or kicking

Who asks afterwards: Parents, the licensing officer, the insurer.

Eldercare & assisted living

Nursing homes, day centres, senior apartments

  • A fall in a corridor
  • Someone on the floor, unnoticed
  • A collision with a trolley or frame

Who asks afterwards: Families, the regulator, the insurer.

Special-needs care

Day activity centres, residential homes, therapy rooms

  • A fall or a knock
  • Hitting or kicking
  • A collision in a doorway

Who asks afterwards: Guardians, the funder, the board.

Home care

Infants with helpers, grandparents living alone

  • A fall with nobody in the room
  • Someone down, with no one there to notice
  • Hitting or kicking, by anyone

Who asks afterwards: The family, before it becomes a hospital visit.

Schools

Playgrounds, corridors, gates, canteens

  • A fall from equipment
  • Hitting or kicking
  • A collision in a corridor

Who asks afterwards: Parents, the discipline committee, the ministry.

Clinics & wards

Waiting areas, bays, corridors, rehab gyms

  • A fall from a bed or chair
  • A patient down and out of sight
  • A collision in a busy corridor

Who asks afterwards: Families, the quality team, the insurer.

One engine. What changes is the setting.

Every deployment runs on the CCTV already in the building. Per site we set which zones matter, who gets the alert and who receives the record.

Talk to us about your setting

Decisions you cannot bolt on later.

Six choices made at the start, because none of them can be added afterwards.

Privacy by architecture

No facial recognition, and what we retain cannot be re-identified. Built into the product, not promised in a policy.

Search, don’t scrub

The archive becomes a question you can ask in English and have answered in seconds.

Local processing

Everything runs on a small unit on site. No cloud round-trips, no footage off-premises, no egress bills.

Sold as a system

Hardware, subscription, on-site setup and training from one team. Plugs into the CCTV you already own.

Human in the loop

SIMIS flags candidates; the people on duty confirm them. It never judges intent or acts on its own.

Built for the conversation after

Structured records and printable reports for the family meeting, the licensing visit and the insurer’s letter.

Recognised by AI Singapore.

AI Singapore (opens in a new tab), the national AI programme, named SIMIS 1st place in the Pan-SEA Developer Challenge, Healthcare Track, 2025.

The SIMIS founders holding the first-prize cheque at the AI Singapore Pan-SEA AI Developer Challenge 2025 prize ceremony
Prize ceremony, Pan-SEA AI Developer Challenge 2025 · Healthcare Track, 1st place

Incorporated in Singapore, April 2026. SUTD ARISE Validate, 2026. Pre-seed; pilot conversations under way with operators in Singapore.

Healthcare people who build AI.

Three founders: regulated healthcare, production AI inside government and pharma, and hardware out of Shenzhen.

Jevin Tan

Jevin Tan

Business & Compliance

Fourteen years in healthcare delivery and regulation across Southeast Asia. Owns commercial, partnerships and the regulatory road map.

Raymond Harrison

Raymond Harrison

AI Engineering

Built production AI for GSK, HTX and URA. Leads the AI behind every SIMIS deployment.

Jenxi Seow

Jenxi Seow

Operations & Hardware

Took products from prototype to volume manufacturing in Shenzhen. Owns the on-site unit, deployment and reliability.

Advisory: AI Singapore mentors and the wider Pan-SEA Developer Challenge network.

Questions we get asked.

Straight answers on what it detects, what it keeps and what it never does.

What does SIMIS detect?

Four kinds of physical event: falls, collisions, hitting and kicking. It flags the moment and the clip for a person on duty to confirm or dismiss. It does not decide what the event means or who was at fault.

Does it recognise faces?

No. SIMIS sees a person, a posture and a movement, never a name. There is no facial-recognition setting to switch on; it was never built. Faces in the clips that are kept are anonymised.

Where does the footage go?

It stays on site. Your recorder keeps recording as it does today. SIMIS processes the feeds on a small unit on your premises and keeps only short clips of flagged events on that unit, under your retention policy. Nothing is uploaded or streamed off site.

Does it work with the CCTV we already have?

Usually, yes. SIMIS connects to most IP camera systems from the last several years and runs alongside the recorder without touching it. Remove SIMIS tomorrow and your CCTV is exactly as it was.

What do staff actually receive?

An alert on WhatsApp, Telegram, email or a webhook into your own system, with the clip and the time. Every confirmed event becomes a written incident record with the review history intact, and daily and monthly summaries follow from those records.

How is it sold?

As a system: the on-site unit, the software subscription, and installation and training from one team. Pilots start with a site’s two highest-risk zones at a fixed price, usually installed in an afternoon.