Simba Khadder

Building AI so I can surf more. Context Engine @Redisinc , fmr Founder & CEO @FeatureformML (acq Redis)

[Simba Khadder @simba_khadder]

I've seen so much success with organizations adoption feature stores. I've also seen lots of confusion. The problem is that the feature store category is broad and ill-defined. In reality, there are three types of feature stores: literal, physical, and virtual.

[MLOps Community @mlopscommunity]

Unpacking 3 Types of Feature Stores // @simba_khadder // MLOps Podcast #265: https://t.co/9VfGOisuuv

[Simba Khadder @simba_khadder]

I've had like 100 companies selling AI SDRs reaching out to me, presumably using their own product. They are the most awkward pseudo-personalized cold email/LI requests. The reach-out doubles as a demo, and it's real bad. IMO, LLMs excel at augmentation, not automation.

[FeatureformML @FeatureformML]

Getting model features into production is hard, but it doesn't have to be! 🌟

[Simba Khadder @simba_khadder]

Excited to announce @FeatureformML 's $5.5m seed round, bringing our total to over $8 million! 🚀

Our virtual feature store fundamentally changes how ML teams define, manage, and deploy model features.

https://t.co/xy8vhdH2w3

[Simba Khadder @simba_khadder]

Excited for our webinar TOMORROW. Now that Feast is unmaintained, many people I talk to are pondering their next steps. Whether you're actively using Feast, considering Featureform, or curious about the evolving landscape of open-source feature stores, this webinar is for you.

[FeatureformML @FeatureformML]

📅 Join us this Thursday, December 7th at 12 PM PT for "OSS Feature Store Comparison: #Featureform & #Feast", hosted by our CEO, @simba_khadder! 🎉

🔗Sign up here: https://t.co/J30KbeoqOp

[FeatureformML @FeatureformML]

🚨 Happening tomorrow, Nov 14th at 8 AM PT! Join us for "Building a #FeatureStore with Featureform & @databricks hosted by @simba_khadder 🎉

Learn how to define & run feature pipelines with Featureform's Python API on Databricks! Q&A to follow.

🎟https://t.co/3XxNuRie9N

[Simba Khadder @simba_khadder]

Had a ton of fun chatting with @CShorten30 on the @weaviate_io podcast! You can listen in to learn more about feature stores and their place in LLM-powered systems.

[Connor Shorten @CShorten30]

Hey everyone! I am SUPER excited to publish our 74th Weaviate Podcast featuring Simba Khadder (@simba_khadder), the CEO and Co-Founder of Featureform (@FeatureformML)!! 🎉

Simba is one of the world's experts on building distributed recommendation systems at massive scale! Simba has brought that expertise into the Featureform virtual feature store, which is packed with exciting technologies!

At a high-level, "feature stores" handle the orchestration of the inputs / features given to machine learning models! There are so many interesting parts to this such as handling the continuous MLOps of data, training, deployment, and monitoring -- as well as system synchronization such as Kafka + Snowflake + S3 + @weaviate_io! ♻️🚀

Simba also presented some really interesting insights on how feature engineering works into the RAG picture! I learned so much from this conversation and I hope you enjoy the podcast!

https://t.co/iACpVydz98

[Simba Khadder @simba_khadder]

@andy_pavlo @CMUDB @FeatureformML Super fun, thanks for having me!

[Andy Pavlo (@andypavlo.bsky.social) @andy_pavlo]

.@CMUDB ML⇄DB Seminar Series Video #5 — @simba_khadder (CEO)
@FeatureformML: Feature Store Architectures & Technical Challenges
https://t.co/pyqinzBxAL

[Simba Khadder @simba_khadder]

@spimescape 5. Getting features in production is painful, and often requires another team.
6. Nailing down point-in-time correctness of features for training, dealing with streaming data, etc. seems impossible

/end, sry for being a bit quick and unorganized with it, happy to chat more

[Simba Khadder @simba_khadder]

@spimescape 3. Features are un-documented, can't be re-used or discovered, and when they are re-used upstream changes often break downstream.
4. Features have no monitoring, ACLs, single source of truth and single source of access

[Simba Khadder @simba_khadder]

@spimescape https://t.co/cd7MkAmw85

This may help, aims to break down the types of feature stores and what they do

[FeatureformML @FeatureformML]

💢😧: "I don't understand, what do you mean the features in our data warehouse aren't available for model serving & inference in prod?"

#datascience #mlops #dataengineering #mlengineer #ml #dataops #featurestores

[Simba Khadder @simba_khadder]

@chrisalbon I love mine. Originally only one person on our team had one, and now everyone does. It feels like a notebook except you can easily organize things. I used to bring a notebook everywhere, this replaced it. wrt iPad, it's kind of like how a Kindle and iPad have different purposes.

[FeatureformML @FeatureformML]

🚀 New Blog Alert! "Feature Versioning: A Key Ingredient To Keeping Your Data Science Projects From Going Off The Rails!" 🚂💥 We're diving deep into the world of feature versioning in #MachineLearning.

Check it out here: https://t.co/JMBoV4vAOn

Time for a 🧵!
1/7

[Simba Khadder @simba_khadder]

@WilsonLimSet @FeatureformML Thanks for having me on, it was a ton of fun!

[FeatureformML @FeatureformML]

🤩 We'll also tackle how to manage features & embeddings for practical LLM use cases!

🔖 A must-attend for the following folks:
- #DataScientists
- #mlengineers
- Anyone interested in enhancing their LLM workflows!

🔗 Be sure to RSVP Here: https://t.co/Qzg3TEqKuw

[Simba Khadder @simba_khadder]

@FeatureformML @AporiaAi @databricks This is gonna be a ton of fun!

[Simba Khadder @simba_khadder]

I have a few invites left for our Executive Cocktail Party at Databricks #DataAISummit tomorrow at 7p. DM me for an invite!