Changsha Ma – Medium

Changsha Ma

Demystifying AI Agents in 2025

Preface: In my work advising organizations on AI solutions, I often see misconceptions about what AI agents can really do. Some dismiss…

Mar 2

Demystifying AI Agents in 2025

Mar 2

Knowledge Base Engineering Guide — Ingestion and Storage

The initial wave of GenAI investments focused heavily on rapid prototyping and experimentation. Many organizations found their early…

Jan 28

Knowledge Base Engineering Guide — Ingestion and Storage

Jan 28

How Complicated Can RAG Queries Get? A Pragmatic Look at Its Limits

In the ever-evolving landscape of data management and analysis, we’re witnessing a significant shift in how organizations extract insights…

Oct 14, 2024

How Complicated Can RAG Queries Get? A Pragmatic Look at Its Limits

Oct 14, 2024

How to Determine If You Will Benefit from Fine-Tuning an LLM

In the fast-paced world of Generative AI, pre-trained large language models (LLMs) have become go-to tools for a wide range of…

Jun 23, 2024

How to Determine If You Will Benefit from Fine-Tuning an LLM

Jun 23, 2024

To Retrieve or Extend? Key Considerations and Research Insights on Using RAG and Long-Context LLMs

A significant development in large language models (LLMs) is the expansion of context windows — the span of text a model can consider at…

Apr 21, 2024

To Retrieve or Extend? Key Considerations and Research Insights on Using RAG and Long-Context LLMs

Apr 21, 2024

How to prepare an instruction dataset to fine-tune LLM?

Fine-tuning large language models (LLMs) on custom datasets is a popular technique to adapt these powerful models for specific downstream…

Mar 4, 2024

How to prepare an instruction dataset to fine-tune LLM?

Mar 4, 2024

Create meaningful representations of data for RAG

In the evolving landscape of generative AI, retrieval augmented generation (RAG) has demonstrated immense promise for open-domain question…

Feb 4, 2024

Create meaningful representations of data for RAG

Feb 4, 2024

Custom metrics for instruction fine-tuning of LLMs

When fine-tuning large language models (LLMs) on downstream tasks, we often rely too much on generic metrics like loss/perplexity. While…

Jan 1, 2024

Custom metrics for instruction fine-tuning of LLMs

Jan 1, 2024

Your RAG Needs Some Scaffolding

Updated in 2025 March: RAG is still not a fully solved problem, and therefore this article is still relevant. I updated the blog to reflect…

Oct 15, 2023

Your RAG Needs Some Scaffolding

Oct 15, 2023

LLM as Knowledge Base v.s. LLM with Knowledge Retrieval

Updates on Feb 2025: LLM with Knowledge Retrieval (aka RAG) has been proven as a more sustainable approach for majority of the use cases…

Sep 17, 2023

LLM as Knowledge Base v.s. LLM with Knowledge Retrieval

Sep 17, 2023

Changsha Ma

Changsha Ma

CS PhD | Seasoned Data Scientist | Connecting dots from research and open source innovations | All of my posts are original.

Following

Help
Status
About
Careers
Press
Blog
Privacy
Rules
Terms
Text to speech