Modular Real-Time Sensor Data Platform

Organization: MAScIR / UM6P
Status: Completed · Defense passed
Timeline: February – August 2026
Repository: github.com/thejokers69/sensor-platform
Portfolio page: mohamedlakssirt.tech/en/pfe


Context

Final-year engineering project delivered in an applied R&D environment at MAScIR / UM6P, supporting research teams managing industrial and laboratory sensor data.

Problem Statement

Research and industrial sensor deployments generate heterogeneous, high-volume data streams without a unified system for visualization, project management, real-time alerting, and reproducible deployment.

Separate web apps per domain (agriculture, pharma, industry) meant duplicated code, fragmented UX, and Excel workarounds for NIR charts.

Solution Summary

A full-stack modular platform for visualizing, managing, and notifying on sensor data in real time — designed for applied research workflows with secure APIs, structured data models, and containerized delivery.

The platform covers the full loop: authenticate → select a project → manage sensors & samples → visualize data → run analytical models → get typed results & alerts.


Key Features

  • Interactive dashboards for live and historical sensor data visualization
  • Project, sensor, and sample management with role-based access
  • Real-time notifications via WebSocket channels for threshold and system events
  • Data export pipelines for reporting and downstream analysis (CSV, Excel, PDF)
  • Internationalized UI for multi-language research teams (FR / EN / AR)
  • AI-assisted chat for querying project context and documentation

Architecture Highlights

CapabilityImplementation
Multi-tenancyProjects with Owner / Editor / Viewer roles
NIR spectra1024-point scans stored as JSONB in PostgreSQL
Mobile syncPOST /api/scans → store spectrum → run model → return results
Model executionDocker script-runner for Python / R / JS models
Real-timeSocket.io WebSocket alerts and live dashboard updates
AuthJWT (+ Google OAuth), audit logging

Technical Stack

Frontend

React, Next.js, TypeScript, Tailwind CSS, shadcn/ui, Recharts, TanStack Table

Backend

NestJS, Node.js, REST API, JWT, Socket.io, TypeORM, class-validator

Data

PostgreSQL (+ TimescaleDB), JSONB spectra, structured exports

DevOps

Docker, Docker Compose, Turborepo, pnpm monorepo


Deliverables

  • Designed a modular frontend/backend architecture with clear domain boundaries
  • Implemented NestJS REST API with JWT authentication and validated DTOs
  • Built project, sensor, and sample management modules with PostgreSQL persistence
  • Delivered real-time data visualization and Socket.io notification layer
  • Added export workflows and structured reporting endpoints
  • Containerized the stack with Docker for reproducible demos and deployment
  • Integrated i18n across user-facing modules
  • Documented API contracts and deployment steps for handover

Metrics (honest, documented)

MetricValue
Domains unified3 (agri / pharma / industry)
NIR pipeline~1024 points per spectrum
Design targetsUp to 50 projects, 10M sensor rows, 1000 sensors/project

These are design targets from the project specification — not measured production KPIs.


TypeURL
GitHubthejokers69/sensor-platform
MAScIRmascir.com
Live demoLocal/Docker only (no public deployment)
Portfolio PFE pagemohamedlakssirt.tech/en/pfe

One-liner Takeaway

One platform for many research projects: from mobile NIR scan to typed chemical results — without Excel workarounds.


On this page