2025-01-27 (updated: 2025-01-27 ) | #deepseek #mixture of experts #r1 model #reasoning model
Last week Chinese AI labs DeepSeek released their latest reasoning model R1, their models are on par with the most advanced models from OpenAI, Anthropic and Meta. This post is about the details of DeepSeek R1 model and the architecture behind it.
2025-01-02 (updated: 2025-01-02 ) | #federated learning #federated transfer learning #horizontal federated learning #vertical federated learning
Federated Learning (FL) is an advanced distributed machine learning paradigm that enables multiple clients such as devices, organizations, or data nodes—to collaboratively train a shared model while maintaining data privacy by keeping the data decentralized.
2025-01-01 (updated: 2025-01-01 ) | #airflow #dag #directed acyclic graph #ml pipelines #workflow orchestration
Happy New Year! 2024 was a busy year for me and I could finally find some time to write. In this post I’ll try to share some details about DAGs and their usage in ML pipelines.
2024-06-01 (updated: 2024-06-01 ) | #deep learning #machine learning #ml model #model #model compression #quantization
It has been a long time since I’ve shared any post so I thought today is the time for that :) In this post I’ll walk you through the details between static vs dynamic quantization which I think might be interesting.
2024-02-18 (updated: 2024-02-18 ) | #aws neuron #inferentia #ml model #ml model optimization
I had a chance to work with AWS Neuron SDK to optimize ML models to enable running inference endpoints Inferentia based instances. In this post, I give some insights about Neuron and how to optimize models using Neuron SDK.
2024-01-01 (updated: 2024-01-02 ) | #deployment #machine learning #ml model #mlops #techniques
In this first post of 2024, I wanted to give some quick insights about the details of MLOps and ML model deployment, and then different techniques used in deployment. MLOps, short for Machine Learning Operations, is a practice within the field of data science and machine learning that brings together the principles of DevOps and applies them to the unique challenges of machine learning model development and deployment.
2023-12-25 (updated: 2023-12-25 ) | #dynamic batch inference #ml #ml model serving #torchserve
Recently I’ve done research on ML model serving frameworks and worked on dynamic batch inference applications. I thought it would be great to share some details here which you might find useful.
2023-12-23 (updated: 2023-12-23 ) | #deep learning #ml model #model #model compression #quantization
Machine learning enables computers to perform tasks smartly by learning from data and instances, instead of just following fixed rules. This is enabled by the vast quantities of data gathered in different sectors and the rapid development in computing power, which collectively bolster the capabilities of machine learning algorithms.
