Internship · Sagemcom2026Industrial Data Analytics Platform
How can anomalies and risks in industrial data be detected, explained and explored in plain language?
An AI-powered analytics platform built during my internship at Sagemcom to detect anomalies, predict risks and monitor industrial data. It combines ML pipelines with drift detection, a multi-agent assistant for conversational analysis and automated reporting, and a React + FastAPI interface with real-time dashboards.
- My role
- AI / ML intern: ML pipelines, multi-agent system and full-stack interface
- Duration
- 3 months (Jun – Aug 2026)
- Platforms
- Web, AI
- Stack
- Python
- LightGBM
- LangGraph
- FastAPI
- React
- Scikit-learn
- SHAP
- MLflow
- LangChain
- pandas
- NumPy
- SciPy
- Groq API
- SQLite
- Docker
- Machine learning
Situation
Internship at Sagemcom (June – August 2026). The goal: an AI-powered analytics platform for anomaly detection, risk prediction and monitoring of industrial data.
Product
- Interactive dashboards, KPIs and analytics on industrial data.
- A conversational assistant to analyse the data, generate reports automatically and work with Excel files.
- Model predictions explained with SHAP.
- Live monitoring of the AI agents, streamed in real time (SSE).
System
- ML pipelines: Isolation Forest and LightGBM, missing values imputed with KNNImputer, statistical drift detection with PSI and Kolmogorov–Smirnov tests.
- Agents: multi-agent system built with LangGraph and LangChain, on the Groq API.
- MLOps: experiment tracking and model management with MLflow; SHAP for interpretability.
- App: React (Vite) front end, FastAPI back end with SSE streaming, SQLite, Docker.