Workshop | Building a Large Language Model from Scratch In-Person / Online
Disclaimer: This session will be recorded. Please let the coordinator know if you do not wish to be in the recording.
Hosted by EEAS Professor Dr Manmeet Singh, this activity guides participants through the end-to-end process of building a Large Language Model (LLM) from scratch, offering a practical and accessible introduction to modern AI systems. Learners begin by exploring the core concepts behind LLMs—including tokenization, embeddings, attention mechanisms, and transformer architectures—before implementing these components in code using intuitive, incremental steps. The activity emphasizes hands-on experimentation: participants construct a simple tokenizer, build a miniature transformer model, train it on a small text dataset, and evaluate its ability to generate coherent language. Along the way, they gain an understanding of how data flows through an LLM, how training optimizes model parameters, and how scaling data or architecture affects performance. The activity also highlights responsible AI considerations, such as dataset selection, bias mitigation, and computational constraints. By the end, participants not only understand the theoretical foundations of LLMs but also experience the full development lifecycle—from model design and training to inference and refinement—equipping them with both conceptual insight and practical skills essential for exploring advanced AI systems.
- Date:
- Wednesday, February 11, 2026
- Time:
- 2:00pm - 3:00pm
- Time Zone:
- Central Time - US & Canada (change)
- Location:
- Helm 3001
- Audience:
- Faculty Graduate Students Undergraduate Students
- Categories:
- Guest Speakers Love Data Week