SmartTodos

Inspiration

We wanted to build an application the used machine learning to help manage our lives. We are super busy, and we cannot spend any extra time figuring out what we are doing when. Computers are great at aggregating data and drawing insights from data, so we thought implementing machine learning for managing lives would be an awesome idea.

What it does

Automatically prioritizes all of your tasks using a dynamic ML model. It will also predict how long each task will take using it's amazing history and engine, to help let you know which tasks will be the easiest/shortest time.

How we built it

We used Vue.js for the frontend, and are running the backend on AWS that consists of Flask, Keras, SciKit Learn, Tensorflow, and Word2Vec models.

Challenges we ran into

Machine Learning on its own does not work too well for pure textual. There has to be significant amounts of feature engineering done on the text before it can be processed through a conventional model.

Accomplishments that we're proud of

We are able to deploy the machine learning model to a (non-beefy) AWS instance, where predictions are served in <20ms. Our vue.js frontend is super high performant. Our ML model achieves high accuracy for the prioritization and time prediction.

What we learned

Flask is great for quick APIs in Python.

What's next for SmartTodos

• Domain specific knowledge: imagine being able to put in natural text, Assignment X for Class Z, and having our ML model predict based on previous personalized history on how long this assignment will take you.
• Model fine tuning exposed. Allow the end user to plugin their actual values for priority and time taken to adjust the model dynamically per user.

Built With

• amazon-web-services
• flask
• keras
• python
• scikit-learn
• tensorflow
• vue