Designing agents

Building, learning, and writing about AI.

The human mind really interests me.

After working as a psychologist for nearly fifteen years, I started craving something different. I wanted a field that moved and changed quickly.

That's when I found AI, before ChatGPT 3 became popular.

I found the top people in the field, studied what they did, and focused my learning.

This made me realize it was time to bring two fields together.

But if you combine two or more skills and get into the top 25% of each, you create a unique mix that sets you apart.

Writing

Notes.

View all

Oct 6, 20265 min read

TF-IDF, BM25 and NDCG

TF-IDF, BM25, and NDCG are core concepts in information retrieval, search engines, and ranking evaluation. TF-IDF and BM25 are algorithms used to score and rank documents based on a search query, while NDCG is a metric used to evaluate how well those search results are ordered

Aug 26, 20264 min read

The LLM Evaluation Lifecycle

Evaluation involves identifying failure patterns and measuring how often they occur.

Evals

Aug 20, 20262 min read

Beating Anthropic's Original Performance Take-Home Challenge

1,361 cycles, down from a 147,734-cycle baseline, with no background in kernel optimization, SIMD, or compiler work.

Agentic Workflows

View all notes

Reach out

ai@andreimoise.com