NEWWhy single embeddings fail for video.Read the post →
    Ethan Steininger's cover

    Ethan Steininger

    Former lead of MongoDB's Search Team, Ethan noticed the most common problem customers faced was building indexing and search infrastructure on their S3 buckets. Mixpeek was born.

    Ethan Steininger

    Recent Posts

    The 3072-Dimension Problem

    The 3072-Dimension Problem

    A 3072-dimensional embedding encodes everything about a video and distinguishes nothing. Decomposing content into named, measurable features, then placing them in a queryable hierarchy, is how multimodal search actually works at scale.

    April 29, 20266 min read
    Why Vector Search Alone Can't Find What's in Your Videos

    Why Vector Search Alone Can't Find What's in Your Videos

    Text-only RAG pipelines miss 80%% of what is in your content. A video contains faces, dialogue, on-screen text, background music, and brand logos. No single embedding captures all of that. The solution is multi-stage retrieval.

    April 26, 20266 min read
    Multimodal Taxonomies: How to Classify Video, Images, Audio, and Text with One Category System

    Multimodal Taxonomies: How to Classify Video, Images, Audio, and Text with One Category System

    Traditional taxonomies classify one content type at a time. Multimodal taxonomies unify classification across every format using embedding similarity the missing layer between raw AI features and structured, searchable metadata.

    April 25, 20265 min read
    Object Storage Comparison 2026: 21 Providers, Real Pricing, and the Gotchas Nobody Tells You

    Object Storage Comparison 2026: 21 Providers, Real Pricing, and the Gotchas Nobody Tells You

    We compared 21 S3-compatible object storage providers across pricing, egress, features, and fine print. AWS S3 costs 15x more than the cheapest alternative for the same workload. Here's everything we found.

    April 10, 20264 min read
    Building a Kalshi Trading Bot with Semantic Search and LLM Extraction

    Building a Kalshi Trading Bot with Semantic Search and LLM Extraction

    How we built an autonomous Kalshi trading bot using the Kalshi API and Mixpeek's video transcription, semantic search, and LLM data extraction no external tools required.

    April 2, 20266 min read
    The Multimodal Data Warehouse: Why Unstructured Data Needs Its Own Snowflake

    The Multimodal Data Warehouse: Why Unstructured Data Needs Its Own Snowflake

    We are drowning in unstructured data — video, audio, images, documents, IoT — but our infrastructure still assumes everything is a row or a vector. The multimodal data warehouse is the missing layer: object decomposition, tiered storage, and multi-stage retrieval pipelines for the AI era.

    March 28, 202610 min read
    ColQwen2 + MUVERA: Multimodal Late Interaction Retrieval That Actually Scales

    ColQwen2 + MUVERA: Multimodal Late Interaction Retrieval That Actually Scales

    We benchmarked every viable approach to multimodal document retrieval on financial tables (ViDoRe/TabFQuAD) and found a combination that hasn't been published before: ColQwen2 + MUVERA. It retains 99.4% of brute-force quality at a fraction of the cost, and obliterates OCR-based search. The Problem Late interaction models like ColBERT and ColPali represent documents as sets of vectors—one per token or image patch. At query time, every query token finds its best-matching document token (MaxSim/

    March 25, 20263 min read
    We Built a Pre-Publication IP Clearance Pipeline. Here's What We Learned.

    We Built a Pre-Publication IP Clearance Pipeline. Here's What We Learned.

    Every major IP enforcement tool finds violations after they're live. We built one that catches them before publication. Here's the architecture, the models, and what we learned.

    March 25, 202611 min read
    I Benchmarked 5 Video Embedding Models So You Don't Have To

    I Benchmarked 5 Video Embedding Models So You Don't Have To

    We tested Gemini, Twelve Labs Marengo, X-CLIP, SigLIP 2, and InternVideo2 on text-to-video retrieval with graded relevance. The results surprised us.

    March 12, 20266 min read
    Gemini Embedding 2 is Live: embed multiple files into one vector

    Gemini Embedding 2 is Live: embed multiple files into one vector

    Google's Gemini Embedding 2 embeds images, PDFs, and text together in a single API call. Here's how we integrated it into Mixpeek's feature extractor pipeline, the production numbers, and where multi-file embedding beats single-chunk approaches.

    March 12, 202610 min read
    Query Preprocessing: Semantic Search With Large Files

    Query Preprocessing: Semantic Search With Large Files

    How we built query preprocessing into Mixpeek's feature_search stage — decompose a 500MB video into chunks, embed in parallel, fuse results. Zero API surface change for callers.

    March 9, 20266 min read
    AI Video Analysis for Sports: Build Automated Highlight Reels, Archive Search, and Performance Analytics

    AI Video Analysis for Sports: Build Automated Highlight Reels, Archive Search, and Performance Analytics

    Sports broadcasters cut 4-8 hour editing sessions to 15 minutes using AI video analysis. Learn how to build automated highlight detection, archive search, and performance analytics pipelines for any sport.

    March 9, 20266 min read
    How Mixpeek runs distributed multimodal ML on Ray: architecture, patterns, and production lessons

    How Mixpeek runs distributed multimodal ML on Ray: architecture, patterns, and production lessons

    We run 20+ ML models in parallel across video, image, and document pipelines. Here's the Ray architecture behind it -- custom resource isolation, flexible actor pools, distributed Qdrant writes, and the lessons we learned the hard way.

    February 25, 20267 min read
    Political Ad Disclaimers: How ZIP+4 Targeting Creates Jurisdiction Conflicts

    Political Ad Disclaimers: How ZIP+4 Targeting Creates Jurisdiction Conflicts

    6,000+ ZIP codes straddle congressional district lines. At ZIP+4 precision, federal, state, and local disclaimer requirements can all apply simultaneously. Here's how multimodal AI solves what static rules engines can't.

    February 19, 20265 min read
    IAB Contextual Classifier: Taxonomies for Videos and Images

    IAB Contextual Classifier: Taxonomies for Videos and Images

    Classify text, images, and video into 700+ IAB Content Taxonomy categories using multimodal AI. Learn how it works under the hood and how to extend it for your contextual targeting needs.

    February 15, 20269 min read