Excited to announce a full suite of new AI-powered tools as we expand beyond email and enable security teams to leverage AI for their end users. Super proud of the entire Abnormal Security team and what we're delivering for our customers!
abnormalsecurity.com/blog/new-produ…

Excited to share our new, incredibly easy to use GitHub action, which will run your evals on every PR, and cache repetitive LLM calls for you automatically using our proxy.

Congrats to Bipul Sinha, Arvind Nithrakashyap, and the Rubrik team on today's IPO (@NYSE: RBRK). Rubrik is on a mission to secure the world’s data, and we at Greylock are proud to be part of their journey since 2015.
#RubrikIPO
cc: Asheem Chandna

greylock.com/portfolio-news…

Most teams building GenAI products today are relying fully on vibes + human review. This works at the start but eventually breaks at scale.

Sharing three approaches we’ve seen engineering teams successfully leverage to get started with automated evaluations:

awesome to see Orb becoming the obvious place to bill for databases— trusted by Supabase, Turso, Neon - Serverless Postgres, Pinecone, YugabyteDB, Materialize, MotherDuck, a bunch more.

The pace of execution and customer centricity with which the Braintrust Data team operates is breathtaking.

It's no surprise that leading AI application companies like Notion, Airtable, Retool, Replit and many more use Braintrust for AI application development.

Every natural language / AI based product needs what Ankur Goyal and his team are building!

It's fun to see that 'vibes eval,' became an industry standard 😅 but while it's useful, we've also gotten good at extracting real-world benchmarks from our data. Also, we use Braintrust Data for continuous evals.

Learn how to build an optimized LLM inference system from the ground up in our new short course, Efficiently Serving LLMs, built in collaboration with Predibase and taught by Travis Addair.

Whether you're serving your own LLM or using a model hosting service, this course will…

excited for us at Greylock to back Richard Artoul Ryan Worl and WarpStream as they reimagine data streaming with a fully compatible Kafka streaming platfrom that's cloud-native and built directly on top of object storage

Fortunate to have partnered with Rishabh Jain, Shreyas and team at Fermat at inception. Building leading AI-enabled commerce platform. 0 to hundreds of brands as customers less than a year after launching. Happy to have Scott Scott Friend and Bain join us on the journey ahead.

Change Healthcare – the nation's largest clearinghouse for insurance claims and payments – has been down for 13 days and counting due to a cyberattack. It's an absolute crisis – doctors, hospitals, and other providers can't get paid. What is a clearinghouse? What is going on? 🧵

.Tuhin Srivastava and team at Baseten are building best-in-class infrastructure to power next gen AI products. Growing incredibly quickly with a high performance, production grade solution.

Last week Saam Motamedi and I co-hosted a dinner with Ankur Goyal at Braintrust Data for engineering and product leaders to talk about the emerging infra stack.

Evals is consistently a top 3 pain point - if that's true for you, Braintrust is here to help

Been digging into Braintrust Data for an LLM eval project. Not often that an API onboarding flow is this seamless.

This is an exceptional intro from Liucija Hostinger on how to do evals. She made this video to share best practices with her team and was kind enough to share it with the world!

It's also fascinating to see the impact of AI on their business -- 40% of support chats use AI!

Absolutely incredible work by Predibase , would be great for GPU-constrained setup.

-------

And if anybody wondering What is LoRAX?

LoRAX works by loading in the fine-tuned “adapter” weights dynamically at runtime. Combining this with an optimized caching and scheduling…

Today we are launching LoRA Land: 25 fine-tuned #mistral 7b #llm that outperform #gpt4 on task-specific applications ranging from sentiment detection to question answering. Try it: pbase.ai/3UJp9lq #llm #opensource #openai
Models are on Hugging Face huggingface.co/predibase

Introducing Fuyu-Heavy, our new multimodal model. Fuyu-Heavy is the world’s third-most-capable multimodal model, behind only GPT4-V and Gemini Ultra, which are 10-20 times larger. In particular, it outperforms Gemini Pro at both MMLU and MMMU...