The Future of Crowd Safety is
Predictive.

Cerberus leverages advanced artificial intelligence to help event organizers monitor, predict, and manage crowd flow in real-time. We build the intelligence so you can build the experience.

Backed By

Redefining Event Security

The Problem

Traditional crowd management is reactive. By the time a bottleneck forms, it is already a safety risk. Current methodologies rely heavily on post-incident responses or manual observation, leaving security teams blind to emerging congestion. This critical delay not only degrades the attendee experience but can rapidly escalate into dangerous scenarios.

Our Vision: Proactive AI

We are developing computer vision and predictive analytics algorithms that integrate with existing camera infrastructure to map crowd density and predict movement before issues occur. By leveraging advanced spatial mapping and continuous machine learning models, Cerberus turns standard feeds into proactive safety heatmaps.

Execution Strategy

01
Apr — May 2026 Current

The "Crash Course" & Simulation Setup

Bridging CV foundations while constructing the virtual sandbox where the MVP will live.

  • Weeks 1-4: Master Object Detection (YOLOv8) & Density Estimation while generating synthetic crowd footage in Unity/Unreal.
  • Weeks 5-8: Divide workloads to architect the simulated environment, establish data pipelines, and set up the PyTorch environment.
02
Jun — Jul 2026

Model Training & Core Integration

Making the model accurate and connecting inference to a usable interface.

  • June: Train the core model on simulated and open-source datasets strictly for head counting and density mapping.
  • July: Build the near real-time backend pipeline and develop the front-end dashboard to display capacity heatmaps and zone alerts.
03
August 2026

Shipping the Simulated MVP

Finalizing a complete, end-to-end software prototype.

  • Weeks 1-3: Stress test the model within the simulation against poor lighting, angle shifts, and heavy occlusions.
  • Week 4: Code freeze. Record high-quality demos of the pipeline ingesting video and outputting analytics for technical documentation.
04
Sep — Nov 2026

The Fall Cohort & Real-World Prep

Leveraging the simulated MVP for resources and bridging the "Sim-to-Real" gap.

  • September: Pitch the UM6P Explorer fall cohort using the working software prototype to secure hardware funding.
  • October: Source physical gear (IP cameras, NVIDIA Jetsons) and map campus installation logistics.
  • November: Collect raw physical footage to analyze model accuracy drop-off and begin fine-tuning.
05
Dec 2026 — Mar 2027

Real-World Pilot & Iteration

Deploying Cerberus into physical reality and handling environmental chaos.

  • December: Initial hardware/software deployment at a controlled campus bottleneck to prove continuous system stability.
  • Jan-Feb: Aggressively retrain the model to handle changing sunlight, weather, and physical occlusions.
  • March: Once baseline density estimation is stable, shift focus to V2 features like localized anomaly detection.

Built by Future Engineers.

Ilyas Rahmouni

Ilyas Rahmouni

Co-Founder & Technical Lead
Amine Chrif

Amine Chrif

Co-Founder & Technical Lead

Help Us Shape the Future

We are actively looking for forward-thinking event organizers, venue managers, and industry experts to become early Design Partners. If you want to pilot next-generation crowd intelligence, let's talk.