May 19, 2025 | NYCUnderstanding and Building
Large Language Models:
From Concept to ApplicationSecure my place

May 19 – Risk Americas Workshop

Understanding and Building Large Language Models: From Concept to Application

CeFPro invites you to a highly engaging and interactive one-day workshop to advance professionals’ understanding of the core components of building large language models.

Registration begins at 9:00 AM and the program concludes at 5:00 PM. The agenda includes allocated time for refreshment breaks, networking opportunities, and lunch.

Due to the intensive and interactive nature of this workshop, seats are limited and will be allocated on a first-come, first-served basis.

Reserve your place today to ensure your participation!

The instructors are volunteers and are just sharing their knowledge with the peers.

What you’ll gain from attending:

  • A foundational understanding of the core components of large language models.
  • Practical insights from real-world use cases, highlighting lessons learned.
  • Hands-on, interactive breakout sessions on scenario development.
  • Guidelines for assessing scenario probability and impact.
  • Networking opportunities with industry peers.

May 19 – Risk Americas Large Language Model Workshop

Agenda


Chris Smigielski,Model Risk Director,Arvest Bank
Chandrakant Maheshwari,Lead Model Validator,Flagstar Bank

8:15am – 9:00am | Registration & Breakfast

9:00am – 10:20am | Introduction Session –

    • Gain a big-picture understanding of LLMs.
    • Understand the full development pipeline, from data to evaluation.
    • Participate in practical exercises for applied learning

Definitions and key concepts.

    • Real-world applications of LLMs in finance institutions.
    • Understanding how a fully functional LLM operates to guide learning.
    • Discussing LLM benefits and limitations.
  • Interactive Activity:
    • Participants brainstorm potential use cases for LLMs in their fields.

Data Pre-processing Pipeline –

  • Overview of Data Pre-processing:
      • Importance of clean, structured, and labeled data.
      • Common challenges in handling large datasets.
      • Data collection and ingestion.
      • Tokenization and normalization techniques.
      • Addressing data bias and ethical considerations.
  • Hands-on Exercise:
      • Participants preprocess a sample dataset using provided tools or scripts.

10:20am – 10:50am | Break and Refreshment –

10:50am – 12:30pm | Attention Mechanism –

  • Role of attention in processing sequential data
    • Evolution from RNNs to attention-based mechanisms.
    • Visualizing how self-attention works.
    • Key concepts: query, key, and value.
  • Interactive Demo:
    • Explore the attention weights of a pre-trained LLM on a sample input.

LLM Architecture –

  • Understanding the Transformer Architecture:
    • Encoder-decoder structure and how they interact.
    • Multi-head attention and feedforward networks.
    • Optimization Techniques:
    • Positional encoding and its significance.
    • Handling long-term dependencies in text.
  • Interactive Visualization:
    • Dissecting the architecture of a well-known LLM like GPT or BERT.

12:30pm – 1:30pm | Lunch Break –

1:30pm -3:00pm  | Training and Fine-tuning –

  • Steps in Training:
    • Model initialization and defining objectives (e.g., cross-entropy loss).
    • Batch processing and backpropagation.
  • Fine-tuning for Applications:
    • Adjusting pre-trained models for domain-specific tasks.
    • Transfer learning and zero-shot learning examples.
  • Hands-on Exercise:
    • Participants fine-tune a small pre-trained LLM on a sample dataset using a provided tool.

3:00pm – 3:30pm | Afternoon Break

3:30pm – 4:50pm | Evaluation Strategies –

  • Key Evaluation Metrics:
    • Perplexity, BLEU, and accuracy.
    • Trade-offs between different metrics.
  • Evaluation Methods:
    • Testing for model robustness, fairness, and bias.
    • Iterative refinement through user feedback.
  • Interactive Exercise:
    • Participants evaluate the performance of a small LLM on a test dataset.
  • Putting It All Together:
    • Design a simplified LLM pipeline using concepts from the day.
    • Teams work on a mini-project (e.g., text summarization or chatbot creation).
  • Presentation:
    • Teams share their solutions and key learnings with the group.

4:50pm – 5:00pm | Closing and Q&A –

  • Recap of the day’s key topics.
  • Sharing additional resources for continued learning (e.g., courses, papers).
  • Open Q&A session for participants to clarify doubts or discuss further applications.

5pm – Conclusion and Wrap Up

Register

Secure Your Place

0Weeks0Days0Hours0Minutes0Seconds
Workshop Ticket

$699
$499*

per person

Expires May 2

*For those representing a financial
institution/government body

PICK THIS TICKET
Workshop + Risk Americas Main Event

$1998
$1498*

per person

Expires May 2

*For those representing a financial
institution/government body

PICK THIS TICKET