ML Features for an IoT Home-Automation Platform (Qolsys)
Company: Qolsys (Oct 2018 – Jan 2020)
Role: Lead Data Scientist
📌Project overview
Led end-to-end development of AI features for a smart home automation and security system, including predictive maintenance, user personalization, and gesture-controlled interfaces.
❗The problem
Sensor data was siloed, unstructured, and reactive- offering no early warnings, intelligent recommendations, or hands-free control for users.
🛠️ What I did
- Designed battery failure prediction using LSTM, RNN, ARIMA models on sensor telemetry
- Clustered sensor behaviour patterns with K-Means, Mahalanobis distance, and Self-Organizing Maps
- Built KNN models to recommend new sensors based on usage, crime data, and regional risks
- Deployed facial and gesture recognition using CNNs + OpenCV for home commands
- Used PySpark for big data processing; implemented ELK stack (ElasticSearch, Logstash, Kibana) for log indexing and visualization
🎯 Impact
- Enabled proactive maintenance alerts and personalised automation
- Reduced system downtime and enhanced home security features
- Strengthened product-market fit by offering futuristic, AI-driven user experiences