#embeddings
13 articles
Hybrid Search: Fusing Dense Vectors and Sparse BM25 Correctly
Combine vector search and BM25 without ruining either — RRF, score normalization, dynamic weighting, and the failure modes that drop precision.
Vector Database Landscape 2026: Pinecone, Qdrant, Weaviate, Milvus, pgvector
A hardheaded comparison of Pinecone, Qdrant, Weaviate, Milvus, and pgvector: architecture, real costs, hybrid search maturity, and when each wins in production.
Multimodal RAG: retrieving images, PDFs, and text together
Build a RAG pipeline that handles mixed corpora of images, charts, and PDFs — covering embedding strategy, retrieval architecture, and production failure modes.
ANN indexes under the hood: HNSW, IVF, and when each wins
A mechanistic walkthrough of HNSW and IVF approximate nearest-neighbor algorithms — with complexity analysis, memory math, and concrete guidance on when each index type wins.
Choosing an embedding model in 2026: OpenAI, Cohere, Gemini, Voyage, and open-source
A decision framework for picking an embedding model: MTEB scores, latency, cost, multimodal needs, and when to self-host — with 2026 benchmarks.
Chunking Strategies That Actually Matter: From Fixed-Size to Late Chunking
Chunking is the highest-variance, least-discussed decision in RAG. The wrong strategy silently kills retrieval quality no matter how good your embeddings are.
Similarity Metrics Demystified: Cosine, Dot Product, and L2
Which similarity metric is right for your embeddings? Cosine, dot product, and L2 explained with real geometry, normalization traps, and model-specific guidance.
Embeddings and positional encoding: how tokens become geometry
How token embedding tables and positional encodings work, why RoPE dominates 2025 models, and what context-window advertising actually hides from you.
How Vision-Language Models Work: Architecture from CLIP to Native Multimodal
From CLIP contrastive pairs to natively multimodal transformers — how vision-language models actually see images, and what that means in production.
RAG from Scratch: Ingestion, Retrieval, and Generation in One Pass
Build a working RAG pipeline end-to-end — parse, chunk, embed, index, retrieve, and generate — understanding every decision before layering on complexity.
What Are Embeddings? Geometry, Meaning, and the Math That Makes Them Work
How embedding models project text into vectors, what semantic neighborhoods mean in practice, and where the math breaks down in production RAG.
Transformers Without the Math
Build the mental model every AI engineer needs: how tokens become vectors, why attention is quadratic, what the MLP layers store, and how depth stacks understanding.
Embeddings and Vector Databases That Scale
What embeddings really encode, how ANN indexes work, which vector database to pick at your scale, and the metadata-filtering pitfalls that quietly destroy recall in production.