Naistro.

I built the systems that schedule music and control venue players.

−80%Scheduling errors

After moving playlist and interruption logic from the frontend into backend services.

3,000+Venues supported

Scale served by the automated track-processing pipeline.

01 / Scheduling & interruptions

Fewer scheduling errors. Music resumed correctly.

Music, adverts, and prayer could compete for the same player. I moved scheduling into backend services, resolved overlapping events, and saved track positions so songs continued correctly after an interruption.

−80%Scheduling errors
NAISTRO / SCHEDULEILLUSTRATIVE EXAMPLE
SAMPLE VENUE / ONE PLAYER
NAISTRO01
POSITION MEMORY00:35
BACKEND SCHEDULER

The song keeps
its place.

NOW PLAYINGMusic
00:35
SAVED POSITION 00:35
01MusicPosition retained
02PrayerPriority interruption
03Music resumesFrom saved 00:35

The advert scheduled during prayer is skipped.

Music resumes from its saved position after a priority interruption.
How I built itRead the engineering details

Giving an interruption priority solves only half the problem. I also had to preserve how much of the track had played, resolve conflicts between interruptions, and align what came next. Time-zone handling kept those decisions tied to the venue’s day.

TypeScript, AWS Lambda & Serverless
Move scheduling decisions into backend services.
Event ordering & interval handling
Resolve overlaps and preserve the played portion of an interrupted track.
SQL & time-zone handling
Assemble the venue’s day from its schedule and local time.

−80% Scheduling errors. After moving playlist and interruption logic from the frontend into backend services.

02 / Audio preparation

Audio processing ran automatically in the background.

Uploaded tracks needed compatible formats and consistent loudness before players could use them. I built queued processing with SQS and FFmpeg, stored prepared audio in S3, and handled failures while preserving compatibility with existing players.

3,000+Venues supported
NAISTRO / AUDIOILLUSTRATIVE EXAMPLE
UPLOAD FINISHED
WAVSample audioOriginal levels & format
JOB 01JOB 02JOB 03
SQS keeps the work queued.
BACKGROUND WORKERDOCKER

FFmpeg

INPUT
OUTPUT
Normalize levelsEncode for playbackStore prepared audio
Waiting for a worker
AACPlayback readyPrepared output / S3
READY FOR
THE PLAYER
Queued audio is normalized and encoded by a background worker before becoming available to players.
How I built itRead the engineering details

Moving processing into a queue let the request finish sooner, but the file was not ready yet. I handled processing and failure states and temporary-file cleanup. Compatibility work also had to refresh cached files and metadata so older players received the updated audio.

Amazon SQS & NestJS
Hand processing work to a consumer outside the original request.
FFmpeg & Docker
Normalize and encode audio in an environment containing the required media tools.
S3 & cache invalidation
Store prepared files and make updated audio available to existing players.

3,000+ Venues supported. Scale served by the automated track-processing pipeline.

03 / Remote player control

Operators could confirm what players actually did.

Operators couldn’t tell whether a player had carried out a command. I built remote controls with device acknowledgements, so sending a command and confirming the player’s response became separate, visible states.

20 secInactive-device detection
NAISTRO / CONTROLILLUSTRATIVE EXAMPLE
VENUE PLAYBACKSAMPLE

One request.
Individual results.

REMOTE COMMANDPause playback
Ready to send
PLAYER APlaying
ConnectedAwaiting playback feedback
PLAYER BPlaying
ConnectedAwaiting playback feedback
PLAYER CUnavailable
No active connectionOthers continue independently
Connected players receive controls independently while unavailable devices remain explicit.
How I built itRead the engineering details

One unavailable device should not erase successful sends to the others. I returned per-operation delivery results and worked on aging connection state so a disconnected player would not look active indefinitely. A sent command still needed to be distinguished from observed playback.

WebSockets
Carry interactive controls between a dashboard and connected devices.
DynamoDB & device heartbeats
Keep track of observed device activity.
Authentication & access control
Keep venue and device operations within the appropriate account scope.

20 sec Inactive-device detection. Time to detect inactive devices through real-time health monitoring.

04 / Playback evidence & monitoring

Playback records survived lost connections.

Lost connections put playback records at risk. I stored records on the device until delivery succeeded, including across restarts, and added monitoring to detect inactive players.

NAISTRO / OFFLINEILLUSTRATIVE EXAMPLE
ON THE PLAYER

Keep the evidence.

PLAYBACK RECORD / 03Actual play timeRetained until acknowledged
PLAYBACK RECORD / 02Actual play timeRetained until acknowledged
PLAYBACK RECORD / 01Actual play timeRetained until acknowledged
FILE-BACKED QUEUE3 records retained
OFFLINE
Saved locally
REPORTING
PLAYBACK EVIDENCE

What actually
played.

Record 01 Record 02 Record 03
Awaiting delivery

Only a successful send removes the acknowledged batch.

RECORD 04A new record stays queued.
Playback records remain on the device until a successful delivery acknowledges the batch.
How I built itRead the engineering details

Removing a record when sending started could lose it on failure. I retained unsent records on the device and removed the sent batch after success. Monitoring had a related timing problem: it needed to allow for delayed evidence before treating an expected event as missing.

Python & file-backed queues
Count actual play time and retain unsent records across restarts.
Scheduled monitoring & AWS
Compare expected activity with device evidence and deliver operational alerts.
SQL reporting
Make playback and catalogue information usable for investigation and operations.
05 / Release engineering

Automated releases reduced deployment issues.

Audio workers needed the right media tools and settings in development and production. I packaged the runtime with Docker and automated build and deployment steps, so application code and its dependencies shipped together.

NAISTRO / RELEASE ENGINEERINGILLUSTRATIVE EXAMPLE
DOCKER IMAGERUNTIME INCLUDED

Code is only
part of the release.

01ApplicationService code
02FFmpegMedia tools
03RuntimeDependencies
REPEATABLE RELEASE PROCESS
01Build imageIn progress
02Run checksWaiting
03Deploy to ECSWaiting
DEPLOYED SERVICECurrent version stays available.

A failed check stops the release before deployment.

A repeatable release carries the application, media tools, and runtime through checks into deployment.
How I built itRead the engineering details

A successful application build did not prove that the deployed worker had everything it needed. Container configuration and environment-specific settings had to travel through the release process alongside the code.

Docker & Docker Compose
Package the runtime and coordinate development services.
GitHub Actions & AWS ECS
Automate builds and deploy the service into its target environment.
MY CONTRIBUTION AT NAISTRO

Naistro took my backend work all the way to the player. I worked on the schedule it received, the audio it used, the commands it followed, and the records it sent back when the connection returned.

Back to selected workExplore CredPalDiscuss an opportunity
Illustrations use explanatory examples, not company screens or private data. Reported impact figures come from my résumé.