-
RAG vs fine-tuning
When to give a model the right documents at question time, when to change its weights, and when a product needs both.
-
Prompting vs fine-tuning
When a better prompt will do, when to train the model on your own examples, and how to tell which problem you have.
-
Vector search vs keyword search
Embeddings against BM25: where each one finds the right passage, where each one misses, and how hybrid search merges the two lists.
-
Agents vs workflows
When to let a model choose each next step, when to fix the path in code, and why most teams should start with the simpler one.
-
MCP vs function calling
What function calling and the Model Context Protocol each do, how they meet in one tool call, and when an MCP server is worth building.
-
LLM-as-judge vs human graders
When a model can grade your AI system's answers, when you need people, and how to check the judge against human marks.
-
Quantization vs a smaller model
When to store a model's weights in 8 or 4 bits, when to switch to a smaller model, and how to check what each one costs you in quality.
-
Mixture of experts vs dense models
How a mixture-of-experts model differs from a dense one in memory, compute per token and serving cost, and how to read names like 235B-A22B.
-
AI engineer vs ML engineer
What AI engineers and ML engineers do day to day, the skills each role needs, how their interviews differ, and where the two overlap.
-
Simple vs log returns
The two ways to measure a return, when each one is right, how to convert between them, and why the gap matters for volatile assets.
-
Quant researcher vs quant developer
What quant researchers and quant developers do day to day, the skills and interviews each needs, and how to choose.
-
Quant trader vs quant researcher
What quant traders and quant researchers do day to day, the maths each needs, how the interviews differ, and how to choose.