Exceptional Capital

New playlist by team Exceptional Capital ❤️🔥 listen on Spotify >> https://lnkd.in/dSzEY4XM

Excited to announce our new tutorial with Pixeltable on building cross-modal video search! Traditional video search stops at metadata and transcripts, but what if you could search videos using images, audio clips, or even other videos? That's the power of multimodal embeddings with TwelveLabs Marengo 3.0 model. 🐎 Here's what makes this game-changing: Set up your video library in minutes, automatically segment videos into searchable chunks, and add an embedding index with a single line of code. From there, you can perform text-to-video search ("find scenes with someone giving a speech"), image-to-video search (match visual patterns), audio-to-video search (find similar soundscapes), and video-to-video search (build recommendations or detect duplicates). All using the same unified API—no complex preprocessing, no manual orchestration. 🔎 Best part? Marengo embeddings work across ALL data types—videos, images, audio, and text. Build product catalogs, search documentation with screenshots, or create content recommendation systems that truly understand multimodal data 🌈 ✔️ Check out the full tutorial: https://lnkd.in/gPnG6j-Z ✔️ Check out runnable Jupyter notebook: https://lnkd.in/g5kgNUMd ✔️ Working with TwelveLabs in Pixeltable docs: https://lnkd.in/gVuaFmF7

Incredible work by team Knowtex!

Streaming CDC ingestion is a powerful capability of the Chronon engine. Need a range scan + aggregation over a production DB at high QPS? These query patterns can be difficult and risky to support in most prod environments. Chronon solves the scalability challenge by combining batch compute + stream processing over the binlog. The result is a low-latency view on your production database that supports any window size with flexible transformations and aggregations, without any load on the underlying table. Updates, inserts and deletes are handled with sub-second freshness. As a bonus, it can also leverage binlog history to produce point-in-time-correct training data. This unlocked a number of high impact use cases at Airbnb, including realtime user facing metrics and many high impact features for critical models. Read more about this (and other input types) here: https://lnkd.in/gXz2Kv7a

Nothing hits better in the morning than coffee, croissants, and conversation. ☕🥐🤝 Come say hi to the Gradient Labs team, alongside our friends at This Week in Fintech, on 18th Feb in London. No agenda, just caffeine and chats. Now’s your chance to head over and ask us anything, from how our AI agent stays compliant in the trickiest support conversations for financial services, to how many dogs show up in our office on a regular basis. (Hint: it's more than one). RSVP here: https://lnkd.in/gTRytWMG

👾 🛸 🪩 🪄 🪐 🎵