What does it take to turn a raw dataset into structured examples that an AI model can learn from?
Join WiMLDS Pune for a practical workshop on dataset curation and preparation. Using a shared sample dataset and starter notebook, participants will work through a cleaning pipeline, domain enrichment, schema design and preparation of examples for supervised fine-tuning (SFT).
What we will cover
- Why curate a dataset? Understanding its purpose and value in AI development.
- Dataset walkthrough: Exploring the sample data, its structure and its challenges.
- Cleaning pipeline: Working through the steps needed to clean and prepare the data.
- Domain enrichment and schema: Adding domain context and structuring the dataset.
- From cleaned rows to SFT-ready examples: Preparing data in a format suitable for supervised fine-tuning.
- Legal aspects to be considered while creating a datasets - copyright, IP, royalty, ownership, guardrails, trademarks, GI, licensing, Traditional Knowledge and Community Rights, Personal Data Protection, Consent and Lawful Use, Confidentiality and Trade Secrets, Database Rights, Cross-border Data Transfer, AI Training and Data Usage Rights, AI Safety and Liability, Consumer Protection and Misrepresentation.
The workshop will conclude with an open discussion on dataset commercialisation models and participants’ questions.
As part of our AI for Bharat initiative, we aim to apply these learnings to developing Indian cultural datasets, with the longer-term goal of working towards publication at a reputed international journal or conference.
Who should attend?
AI/ML practitioners, data engineers, software developers, researchers and others interested in the practical engineering behind AI datasets.
Come prepared
This is an interactive workshop. Please bring your laptop. A starter notebook, sample dataset and dependency instructions will be shared beforehand. Participants are expected to install the required dependencies and download the dataset before attending.
Build practical skills in engineering the data that models learn from.
Full text on Meetup
.