At Replit, we extensively use Spark on Databricks for all our training data work. We run highly customized transformations on code like parsing, deduping, PII redaction, code filtering, tokenization, and more, written with low-level Spark primitives. High quality data is key.

POC with big proprietary models, deploy to production with smaller specialized models. Replit a great example of smaller specialized model offering improvements over larger proprietary ones at a better price

Back in June we . showed that our LLM Foundry training stack runs seamlessly on AMD MI250 GPUs. Today, I'm happy to share that we've scaled up to 128xMI250, with great multi-node performance!

An organization is its people. DM me or email me at [email protected]

AI should be decentralized as much as possible. Open source is a great step toward this. Crypto should be decentralized. Self-custodial wallets and protocols help this. Permissionless access creates innovation instead of gatekeepers. No single entity to capture.

Meet our new AI, #DBRX DBRX is an advance in what language models can do per $. These economics will have profound impacts on how AI is used, and we've built this to democratize these capabilities! It's the best open model in the world. It closes the gap to closed models in a

New blog post on DSPy, a recent #llm ops innovation led by researchers from Databricks co-founder Matei Zaharia's lab, co-authored by Dan Pechi and Arnav Singhvi.

Perplexity is actually just better than Google for clear well cited answers. The quality of results and speed to answer is significantly better.

Databricks to acquire Tabular (now part of Databricks), a data platform from the original creators of Apache Iceberg. Together, we will bring format compatibility to the lakehouse for Delta Lake and Apache Iceberg databricks.com/blog/databrick…

NO GOALS FOR YOU!!!! Christian Lomeli is your TST 2024 Golden Glove Winner! What a tournament, Christian👏🧤

Lynx is a new hallucination detection model for #LLMs that is especially suited for real-world applications in industries like healthcare and fintech. PatronusAI trained Lynx on Databricks Mosaic AI using Composer, our open source PyTorch-based training library.

Great post on how AI can be engineered "brick-by-brick" to get the best results, including through the design of compound AI systems. The teams that figure out how to get the right data pipelines and system design are the ones getting transformational results.

Per insider sources: Inter Miami will receive 37 minutes of extra time if they aren’t winning at the end of regulation tonight

o3 or its checkpoints have been available to OpenAI developers and PMs internally for a while now. With virtually unlimited PhD level AI developers and capitol, why haven’t they shipped more world class software? Their other product releases were less exciting than google’s, no?

OpenAI's high percentage of revenue from consumer apps could push them to commoditize their models via open source. Ironically, this would turn ChatGPT into a "thin model wrapper”