From Feature Store to Feature Platform

Congratulations to Jack Antonoff for his third consecutive Grammy for Producer of the Year.

You too can get yourself on a production winning streak like that by attending our AI in Production conference on the 15th and 22nd of February.

Headlining talks performed by superstars from Netflix, Notion, Perplexity AI, Uber, Microsoft, Google, LinkedIn, Reddit, and other great companies. All there to help solve your problems with cost optimization, latency, trust in outputs, and debugging. Plus workshops to streamline setting up your projects, hassle-free.

Rumors of me wearing a designer dress for the red carpet can neither be confirmed or denied at this time.

Register here!

Powering MLOps: The Story of Tecton's Rift // Matt Bleifer and Mike Eastman // MLOps Podcast #209

To continue the Grammy theme, and because we chat a little about music in this episode, I'm linking a song for the items this newsletter.*

For this podcast, it's "Rift" by Phish.

Rift is the name of their Python-based compute engine. It can run Python and Pandas transformations, which allows for easier feature development and increased iteration speed.

We also talk about the development of Tecton from feature store to feature platform, the potential of real-time feature pipelines, organizational architecture, and their work integrating DuckDB.

So ducks, but no phish.

Video || Spotify || Apple

* inclusion of a song does not necessarily mean endorsement or reflect my musical tastes. It's very difficult to link in Gregorian chants...

Don't miss out on an engaging fireside chat featuring Rajat Arya, where we'll dive into the complexities and challenges involved in developing an advanced ML platform.

This conversation offers a chance to gain direct insights into the nuances and obstacles faced while crafting state-of-the-art ML platforms.

Secure your spot to participate live, pose your questions, engage with fellow enthusiasts, and ensure access to the session recording.

📆Tuesday, February 27 (10 AM PST)

Register here!

💬 Join our AI and MLOps Slack Community: slack.flyte.org

⭐️ Start Building With us on GitHub: github.com/flyteorg/flyte

Gen AI Buy vs Build, Commercial vs Open Source // Nayur Khan // Ilona Logvinova // Mohamed Abusaid // QuantumBlack Roundtable

It has to be a pairing of "Can't Buy Me Love" by The Beatles, and "Build Me Up Buttercup" by The Foundations. Both great songs with their own unique merits... you can see where this is going can't you?

This was a great roundtable about the decision-making processes involved when deciding whether to build or buy for Gen AI solutions.

Key points when comparing enterprise models versus open ones were discussed, plus the significance of data security, and the importance of considering cultural and behavioral differences when deploying globally.

At the heart of the discussion was the need to understand an organization's needs and problem-solving demands when deciding when to build or buy.

So a better match might be "Know You Better" by the Black Pumas.

Video || Spotify || Apple

Vexed by Vector Databases? Mad with Multimodal Semantic Search?

Umar Igan has got the course for you!
Introduction to Vector Databases and Multi-Modal Semantic Search.

Learn all about vector databases, semantic search, embeddings, containerization, deployment considerations, and best practices.

Plus, to apply the theory you’ll be walked through the practical aspects, demonstrating the development of a multimodal search on a dataset.

All for the discounted price of $99!

And talking of discounts, the last day of the 40% discount for the Introduction to Q&A Systems with LLMs course is 14 Feb.
The perfect way to treat that special someone in your life this Valentine's Day! 💘

💡Job of the week

Senior engineer // Nomic AI (New York, US)

As a Senior Engineer at Nomic you will work work with the Atlas team to develop, deploy, and scale automated methods for interacting with internet scale size unstructured datasets of text, images and embeddings.

You will:

Manage the Atlas HTTP API layer, integrating it with user interfaces.
Ensure the architecture of Atlas is scalable, efficient, and easy to maintain.
Develop innovative embedding space interactions for Atlas.
Enhance Atlas algorithms for processing vast datasets.
Collaborate with platform engineering to support business user requirements.

Requirements

Strong interest in building resilient, efficient and scalable distributed systems.
Evidence of exceptional ability to learn quickly and bias towards building.
Strong ability across the stack with specific deep knowledge in Python, FastAPI, NextJS, and experience on AWS.

Babs' Demo // Temi Babs // IRL Meetup #63 Lagos

For this, I'm going to go with "The Message" by Grandmaster Flash and the Furious Five, as the focus is on optimizing the messages placeholder.

In this demonstration, Temi Babs discusses the LangChain framework and its integration with OpenAI's LLM to create rich responses for users.

He looks at streaming token by token, and using JSON responses to enhance the user experience. He also delves into the use of Pydantic for data structuring and the potential for fine-tuning the GPT-3.5 and GPT-4 models.

Watch it here!

Become The Maestro Of Your MLOps Abstractions

Wilder Wood's "Maestro (Tears Don't Lie)" comes to mind for this one. I'm sure plenty of tears have been shed trying to manage abstractions, but no more after reading this!

As ecosystems become more complex the need to develop abstractions to integrate various components also grows. This blog serves as a guide for building strong MLOps platforms. It highlights the key design patterns necessary for developing straightforward yet impactful abstractions. These abstractions enable easy integration of new components and technologies.

So no more tears once you become the Maestro. No lie!

With thanks to Médéric Hurier for the contribution.