Mini machine learning platform for arrival times
Inspired by publicly reported work at Uber
Build a small ML platform (features, training, registry, serving, monitoring) and run an ETA model on it.
The brief
Uber has described Michelangelo, its platform for building, deploying and monitoring models, and DeepETA, which runs on it. Build a small platform of your own and run an arrival-time model on it end to end: a feature pipeline, repeatable training, a model registry, an online prediction service and monitoring that catches drift. Good looks like: one command to retrain, versioned models, an API with stated response times, a dashboard of live error and drift, and a written design doc.
Read the case first
What Uber has said in public about this kind of work, on its own pages:
- DeepETA: How Uber Predicts Arrival Times Using Deep Learning (Uber Blog) ↗
- Meet Michelangelo: Uber's Machine Learning Platform (Uber Blog) ↗
Make it your own
Run a second, different model on the same platform (fare prediction, demand by area, or something from your branch) to prove the platform is reusable, and size it for a startup's budget, not Uber's.
AITeky has no link with Uber, which has not reviewed or endorsed this brief. The brief only draws on what the company has published, uses open data rather than its data, and is a learning exercise: don't present your project as Uber's work or system.
What you will build
- Build a feature pipeline from trip records
- Make training repeatable and track experiments
- Register and version models
- Serve predictions through an API with a latency target
- Monitor error and data drift, and alert on it
- Write a design document and a runbook
Free, open data to build it with
Sign up free to get the data and links
It takes a minute with your email. AiTeky Pass (₹999/month or ₹9,999/year) is free for your first 30 days.