We instrumented elders’ apartments with multiple IoT sensors and developed a software solution, marrying multiple open-source software packages, and stored all the data in a large graph datastore.  From this, we inferred human Activities of Daily Living (ADLs), and used Latent Dirichlet Allocation (an NLP topic modeling technique) to identify human behaviors of elders in their apartments.  We also experimented with principal component analysis, anomaly detection, Bayesian graphical models and Recurrent Neural Networks, using Python, Jupyter, sci-kit learn, TensorFlow and Spark.  Our novel solution allowed elders to remain in their own homes longer.  One couple remained in their own home an additional 7 months.