I design AI that people trust and actually use.
For the past two years I've designed an AI-powered workforce platform for construction and facility management teams. Whatever the sector, my job is the same. Take what the model can do, and make it simple enough to use and clear enough to trust. In total, I have 5+ years of design experience.
Four AI features I designed. All four are live.
Shipped and live on construction sites. See how each one works below.
- 01Live
Voice notes
Speak in any language, get structured English notes.
Try it - 02Live
Face registration
Register once, guided by the camera and short voice prompts.
Try it - 03Live
Time & attendance
Face check-in that adds up to timesheets and payroll.
Explore - 04Live
Toolbox Talk Analyzer
Ticks off a safety talk's checklist while it happens.
Try it
Speak in any language. Get structured notes.
Workers talk instead of filling in forms. The AI writes it up in English, and they check it before it's sent.
Recordings from the kinds of people the product serves. Simulated
Speak a work update, a request or a problem. Live in your browser
Uses your browser's built-in speech recognition (Chrome, Edge, Safari). Nothing is stored.
No mic? Type what someone on site might say, in any language. Live AI
- Listen
- Understand
- English
- Structure
- You review
Letting people see and fix what the AI understood raised trust scores by 60% in our user survey.
No fields, no categories. If reporting takes thought, people don't report.
It made transcription reliable for crews who don't report in English.
Register a face in one go.
The camera gives one instruction at a time and checks the photo live, so it passes before it's taken.
Ready when you are
The camera runs on your device. Nothing is uploaded or saved.
- 01Find your face
- 02Fit it in the oval
- 03Hold still
- 04Blink once
- 05Registered
Overtime is worked out from these check-ins and goes to the supervisor and HR to approve. Simulated
Three instructions at once is noise. The screen only asks for the next thing.
Only in this demo. In the shipped version a supervisor is present at registration, which covered that risk.
Supervisors can mark attendance by hand, so nobody loses a day's pay.
Check in with your face. The rest adds up.
Day stats, project hours, timesheets, payroll and leave all follow from one check-in. Pick a part to see it.
Checks in and out with their face, on site.
Matches the registered face and checks the site's GPS geofence.
Guided registration, clear feedback when a match fails, and a manual fallback for supervisors.
Sees check-in, check-out, total, OT and break hours.
Works the hours out, so nobody adds them up by hand.
One screen with only the numbers that matter, readable at a glance.
Their hours are tied to that day's project.
Records the project with each check-in, so time lands on the right job.
The project shown at check-in, so mistakes surface that day, not at month end.
Gets a monthly timesheet and sees what their pay is based on.
Rolls daily records into a timesheet that feeds payroll.
A timesheet they can check line by line, to cut payday disputes.
Applies for leave by voice, in their own language.
Turns the request into a leave application and sends it for approval.
A simple flow with a clear status, so nobody chases a supervisor for an answer.
Live coaching for a safety talk.
The AI listens and ticks off each checklist point as the supervisor speaks. Press Start talk to hear one.
Press Start talk. The script plays in the supervisor's voice, as text.
- Inspect your harness before useTo go
- Clip to an approved anchor pointTo go
- Keep three points of contact on laddersTo go
- Check guardrails and edge protectionTo go
- Tether tools when working above peopleTo go
- Stop work in strong windTo go
- Know the rescue plan if someone fallsTo go
- Covered
- Partly
- Still to go
Cover these tomorrow, or add a quick word before work starts.
A supervisor felt red "missed" points marked him wrong in front of his crew. The missed list now waits for the summary.
A climbing percentage would pull the supervisor's eyes off the crew. The shipped app shows the result after the talk. This demo shows it live so you can see the matching.
One photo after the talk matches the crew. Anyone not matched is tagged by the supervisor.
Eight case studies, from AI features to websites and mobile apps.
See all projects →