Skip to main content
Industrial computer vision · LLM applications · MLOps

AI systems that hold up in production.

We build computer vision, LLM and MLOps systems for industrial and operations teams — and keep them working after launch.

Book a callSee the work

30-min call · no obligation · you talk to the engineer

  • Vision systems deployed at an industrial AI lab
  • Live RTSP cameras, real-time GPU inference
  • Led by the engineer who builds it

This black hole is ray-traced live on your GPU — the same physics-first engineering we ship.

WORK

Systems our lead engineer built and deployed

Built and deployed at an industrial AI lab · client anonymized

PPE-detection demo — helmet and hi-vis vest detected on one worker, missing helmet and vest flagged on another, each with a confidence score.
Vision · Safety

PPE and danger-zone monitoring

Challenge
Safety compliance relied on manual spot-checks that miss violations between rounds.
What we built
PPE and danger-zone detection on live RTSP feeds, with per-person tracking and automated alerts.
Result
Continuous monitoring on every camera, every shift.
Related service: Safety vision
Granulometry demo — rock fragments measured from an overhead camera with millimetre width and height and keypoint detection, against a checkerboard calibration target.
Vision · Measurement

Automated granulometry

Challenge
Rock sizing on the line relied on slow, inconsistent manual sampling.
What we built
An overhead camera with calibrated detection that measures each fragment to millimetre accuracy.
Result
Automated, continuous granulometry that replaces manual sampling on the line.
Related service: Custom CV systems
Vision · Inspection

Conveyor-belt inspection

Challenge
Belt faults were caught only after escalating into downtime.
What we built
Continuous anomaly detection on moving belts, surfacing faults on an operations dashboard.
Result
Defects flagged early — before they stop the line.
Related service: Custom CV systems
MLOps · Platform

Streaming core

Camera ingest and swappable models, no pipeline rebuilds.

  • RTSP / HTTP ingest
  • Hot-swappable models
  • GPU inference
  • Drift monitoring
Related service: Video AI pipelines
PROBLEMS

Does one of these sound like you?

Vision

You've got cameras, but no insight

Your CCTV records footage nobody watches. We turn it into live detection.

See safety vision
LLM

Your team keeps searching the same documents

Answers hide in manuals and tickets. We build assistants that cite their source.

See LLM applications
MLOps

It worked in the demo, then died in production

Real conditions break models. We test against yours, then watch for drift.

See how we deliver

Recognize one? That's what a first call is for. Book a call

SERVICES

Three practices, one standard: it has to run in production.

02

LLM applications

Assistants that answer from your documents and data, tested and guarded.

RAG · Agents · Evals · Self-hosted modelsExplore LLM applications
03

MLOps

Serving, monitoring and retraining that keep models fast and accurate.

Triton · TensorRT · Docker · MonitoringHow we deliver
PROCESS

From an open question to a system you can operate

  1. 01

    Scope and feasibility

    We check your data and goal, and tell you if AI is worth it.

  2. 02

    Prototype on your data

    A working model on your own data, tested on the hard cases.

  3. 03

    Production deployment

    Containerized GPU services, built to fail safely.

  4. 04

    Monitor and improve

    Dashboards and drift alerts catch problems before your users do.

EXPERIENCE

Where we've built and learned

Teams we've built with and learned from:

Zakariae Mrabet, founder of Apollume
Zakariae MrabetFounder and lead AI engineer42 / 1337 graduate · Arm Ambassador
ABOUT

An AI engineering studio, led by the engineer who builds it.

An AI engineering studio in Morocco, working worldwide. We ship models that run every day.

More about Apollume
BLOG

Field notes

All posts

Have a problem worth automating?

30 minutes. A straight answer on whether AI fits your problem.

Book a callYou talk to the engineer who'd build it
What to expect on the call
  • A look at your setup and problem
  • A straight answer on whether AI fits
  • No hard sell