-
Notifications
You must be signed in to change notification settings - Fork 1
Charter
Shree Chaturvedi edited this page Jun 13, 2026
·
1 revision
- Organizations and researchers increasingly rely on data science to solve complex problems (e.g., satellite failure prediction, drug discovery).
- Teams face persistent barriers:
- Time-consuming: data preparation and feature engineering take weeks.
- Limited accessibility: existing AutoML tools don’t easily incorporate experts’ domain knowledge.
- Lack of trust: models often act as “black boxes,” providing little transparency for decision-makers.
- Net effect: slower innovation, reduced confidence, higher costs.
- Build a cloud-based AutoML platform that:
- Streamlines the process from raw data → deployable model.
- Incorporates expert knowledge directly into the workflow, making the system smarter and more relevant.
- Provides transparency by explaining how models reach their predictions.
- Balances automation with human oversight: fast defaults for speed + deeper controls for experts.
- Enable seamless data ingestion & preparation.
- Provide interactive tools for analysis and visualization.
- Automate feature engineering & model training while allowing expert guidance.
- Offer one-click deployment & monitoring.
- Deliver clear, trustworthy explanations of results.
- Demonstrate success on one real-world dataset (e.g., aerospace, biotech).
- End-to-end workflow demonstrated on at least one real dataset.
- Clear evidence that the tool reduces effort and time for data scientists.
- Positive feedback from users on usability and clarity.
- Models with explanations that increase confidence in decisions.
- Technology risks: complex data integration; ensuring system reliability.
- Process risks: balancing workload and timelines within the semester.
- Adoption risks: designing a tool powerful for experts and simple for newcomers.
- Ayush Yadav
- Shree Chaturvedi
- Zarif Fida Chowdhury
- Shree Chaturvedi