Inspiration
Estimating work is a hard challenge for human beings. Google Search has 345 000 000 hits on why estimates always go wrong. What if AI helped?
What it does
Farseer AI accurately predicts how long would it take the team to complete an issue. It analyzes the team delivery patterns, issue content and other parameters on more than ten different dimensions and continuously learns from team progress. It trains a real machine-learning model directly in Forge and uses it to predict completion times. The model is updated on-the-go, and every new closed issue contributes to the accuracy fo the predictions.
How we built it
Implemented a machine learning model that trains from the issues and sprint data and then produces predictions as a Forge app. The model training is done in real-time, it always operates on the freshest data and learns from the team behavior.
Challenges we ran into
Saving the model in Forge Storage is not possible due to the Storage API limits.
Accomplishments that we're proud of
Using a real machine learning model for solving a meaningful hard problem that helps teams be more efficient. Solving a problem that gave nightmares to teams around the world. It's a simple, efficient, and truly disruptive app.
What's next for Farseer
Additional features, learning from custom fields specified by users, more granular settings menu for adjusting estimation behavior to the team's ways of work.
Log in or sign up for Devpost to join the conversation.