← All work

AMS Tablet

Two workers in white uniforms, one in a hair cover, facing the AMS wall unit at a check station while another person points at its screen; faces are blurred
The tablet in use at a check station. Faces blurred.

Work · Automated Monitoring Solutions, Aktau · Nov 2024 – Sep 2025

AMS Tablet

A pre-shift checkpoint tablet for industrial sites in Kazakhstan.

Before a shift, workers on an industrial site pass a checkpoint. AMS put a two-part wall unit there that runs four checks in one stop: identity by face recognition, temperature, breath alcohol and SpO2.

Each unit is an edge device that reports to a cloud dashboard, and an admin dashboard and mobile app hold the attendance and health logs. I worked on it as a full-stack and hardware developer.

Details

Role
Full-stack and hardware developer
Company
Automated Monitoring Solutions (AMS)
Location
Aktau, Kazakhstan
Dates
Nov 2024 – Sep 2025
Checks
Identity, temperature, breath alcohol, SpO2
Built with
Raspberry Pi, AWS, Docker, Portainer, YOLO, Frigate

The checks

Four checks, one stop

One wall unit in two halves: a screen with a camera and a temperature sensor in its top edge, and beside it a breathalyser port above a pulse sensor. Each part answers one check.

The unit's instruction card, cropped to its drawing: the camera and the temperature sensor in the top edge of the screen half, the breathalyser port and the pulse sensor on the right half, each circled in red and labelled in Russian
  1. Camera · face recognition
  2. Temperature sensor · remote
  3. Breathalyser · Dingo B02
  4. Pulse sensor · pulse oximetry
The unit's own instruction card; its labels are in Russian.

  1. Identity

    Camera · face recognition

    The first question at a checkpoint is who is standing there. The camera in the unit's top edge takes the worker's face, and face recognition answers it: the deck's component page names AWS face recognition for that job. The readings that follow then have a name on them, and the admin dashboard and mobile app keep that name, with the time, as attendance.

  2. Temperature

    Temperature sensor · remote

    The second reading needs no contact. The deck lists remote temperature measurement among the unit's components, and the instruction card marks the temperature sensor in the same top edge, right beside the camera and facing the same way. The value goes into the health log with the others, under the name the identity check found.

  3. Breath alcohol

    Breathalyser · Dingo B02

    The right half of the unit is built around one part: the round port in its middle is the breathalyser, the largest opening on the front. The deck names the sensor behind it among its components, a Dingo B02 alcohol sensor. The worker breathes into the port, and the result joins the identity, the time and the temperature in the same pre-shift record.

  4. SpO2

    Pulse sensor · pulse oximetry

    The fourth reading takes a fingertip on the round sensor below the breathalyser. The deck's component page calls it pulse oximetry: it reads SpO2, the share of oxygen the blood carries. With that the record is complete, and it goes to the admin dashboard and the mobile app, which hold the attendance and health logs for the site.

On site

On the wall

The unit in the field: on walls at site checkpoints, next to turnstiles and doors. I installed the cameras and sensors on site. Two of the installed tablets show ESG, which is a client company's brand, not the product's name.

The black two-part unit fixed to a white wall, cable running down from it, a Raspberry Pi desktop on its screen and the breathalyser port on the right half
Installed on a wall, a Raspberry Pi desktop on its screen.
A slim wall tablet showing 'AMS' on a pink screen, mounted on a glass partition beside a checkpoint turnstile
Installed at a site checkpoint, beside the turnstile.
A finger pressing the button 'Открыть дверь', open the door, on a white wall tablet whose screen shows the ESG company logo
A client-branded tablet: ESG is the client's brand. The button reads “Открыть дверь”, open the door.

Behind the tablet

Edge to cloud

The two-part unit is built on a Raspberry Pi (its desktop is on the screen in the first wall photograph above), and I did the assembly, wiring and soldering. The units are edge devices: each one reports to a cloud dashboard through an edge-to-cloud API, and the cloud side is deployed on AWS with Docker and Portainer.

I also set up person detection on cameras with YOLO and Frigate. In the feed below, a fabrication workshop, each person found is boxed in red; zones are drawn over the camera image, here a night view in infrared.

  1. Edge The unit A Raspberry Pi with a camera, a temperature sensor, a breathalyser and a pulse sensor.
  2. Link The API An edge-to-cloud API: every unit reports to the cloud.
  3. Cloud The cloud The cloud dashboard, and a deployment on AWS with Docker and Portainer.
  4. Admin The logs The admin dashboard and mobile app hold attendance and health logs.
A camera view of a metal fabrication workshop in the Frigate interface, two people outlined in red boxes among steel frames and a forklift
Person detection in Frigate: a workshop camera, each person found in a red box.
Frigate's Masks and Zones screen: a teal zone drawn across a night infrared camera view of a yard
Setting a zone on a camera, on its night infrared view.
The admin mobile app's main screen: buttons to add employees and see their status, and an entry list of dates and times, names blurred
The admin mobile app: entries by date and time. Names blurred.

Pitch deck · 4 pages, in Russian

The deck

Pitch deck, the product: a render of the two-part AMS wall unit beside the title 'ams system' and four icons for temperature, alcohol, pulse and face checks
Pitch deck, the problem: four statements in Russian about manual pre-shift medical checks of workers
Pitch deck, the components: face recognition with Amazon Web Services, remote temperature, pulse oximetry and a Dingo B02 alcohol sensor, drawn around a render of the unit
Pitch deck, the admin mobile app: a feature list in Russian beside the app's attendance screen, names blurred
  1. The product
  2. The problem: manual pre-shift checks
  3. The components
  4. The admin app

My part · Nov 2024 – Sep 2025

What I did

  1. Software The full-stack app.
  2. Hardware Raspberry Pi assembly, wiring and soldering.
  3. On site Camera and sensor installs.
  4. Cloud Deployment on AWS with Docker and Portainer.
  5. Edge to cloud The API the units report through.
  6. Cameras Person detection with YOLO and Frigate.