Loading Events

« All Events

  • This event has passed.

LLM & Adaptive ML

January 3 @ 6:30 pm - 8:30 pm IST

# MLOPS for Large Natural Language Models
The latest trend in AI is that larger natural language models provide better accuracy; however, larger models are difficult to train because of cost, time, and ease of code integration. To advance large model training by improving scale, speed, cost, and usability for model developers worldwide, we will demonstrate how DeepSpeed library open source can work to publish larger natural language models.
DeepSpeed can efficiently train models with 100–200 billion parameters up to 10 times faster than state-of-the-art via a memory optimization system called ZeRO (Zero Redundancy Optimizer). ZeRO is a parallelized optimizer that greatly reduces the resources needed for model and data parallelism while massively increasing the number of parameters that can be trained. Researchers used these breakthroughs to create Turing Natural Language Generation (Turing-NLG), one of the largest publicly known language models at 17 billion parameters.
## About the Speaker: Efrat Elhay
Cloud Solution Architect at Microsoft specializing in Azure PAAS solutions and open-sources libraries,
an expert in DevOps/MLOPs solutions, with vast experience working with startup and innovation centers.
Before Microsoft Elhay was DevOps/MLOPS leader for advanced analytics in intel that provides all ML/AI cross-intel companies.
# Continuous Integration and Deployment of Adaptive Machine Learning Models for Email Phishing Detection
This talk presents a novel method for calibrating a machine-learning model for detecting email phishing attacks. Our approach leverages big data technologies like Apache Airflow and Apache Spark to process large amounts of classification data in near real-time, and adjust the thresholds of the model to improve its performance. This not only helps us to provide consistent precision across thousands of different customers, but also enables us to automate and streamline the ML model training and deployment process. We will demonstrate the effectiveness of our method through real-world examples, and discuss how it can improve collaboration and coordination between different teams involved in the MLOps workflow.
## About the speaker: Mordechai Worch
Mordechai is the Data Group Lead at Ironscales, where he manages a team of data engineers and machine learning engineers, and is responsible for driving the development and deployment of advanced machine learning models. In addition to his professional expertise, Mordechai is also a volunteer organizer for the Tel Aviv chapter of MLOps.Community, a global community of practitioners and experts focused on the intersection of machine learning and operations. With his unique combination of technical skills and community leadership, Mordechai is well-positioned to share insights and best practices on the challenges and opportunities of MLOps


January 3
6:30 pm - 8:30 pm IST
Event Category:


Totseret ha-Arets St 7
Tel Aviv-Yafo, 6789104 IL
+ Google Map


Tel Aviv-Yafo MLOps Community
View Organizer Website