The Data Stack Show Newsletter Edition 023
LLM fine-tuning is like preparing for an algebra test
👋 Hi all,
Rishabh Bhargava isn’t the first guest we’ve had on the show to talk about LLMs. Gradient co-founder, Mark Huang joined us to discuss AI in the enterprise, Brendan Short brought the knowledge on how LLMs are transforming sales, and Vectara’s Amr Awadallah told us how generative AI is transforming search. And those are just a few notable examples.
But on his show last month, Rishabh gave us the best explanation of fine-tuning I’ve heard yet. He compared model training to preparing for an algebra test. The analogy brings the power and possibility of generative AI into full focus. “...If you truly wanted to ace the test, you wouldn't just show up with a textbook. You'd spend the previous week actually preparing…” Something about knowing how much we can improve our own performance with dedicated effort drives the point home. Listen to the show for the full quote, and check out the rest of our lineup from last month.
🌯 The February Wrap
Here’s what you missed (if you missed it) on The Data Stack Show last month:
The Fundamentals of Event-Driven Orchestration and How Generative AI Is Shaping Its Future with Viren Baraiya (Twitter) of Orkes
Why you should listen – To learn all about orchestration and microservices. Viren takes orchestration out of the typical “orchestration tool” box to get at the fundamentals. He looks at orchestration from the perspective of data and software engineering, then expands it to teams like product to explore different applications. He also talks AI (of course) and tells us how LLMs are impacting the orchestration world.
AI-Based Data Cleaning, Data Labelling, and Data Enrichment with LLMs Featuring Rishabh Bhargava (Twitter) of refuel
Why you should listen – For a look into some AI use cases that cut through the hype and get down business. Rishabh introduces the idea that humans could write instructions once, and then have machines do all of their data cleaning, labeling, and enriching for them. Then he tells us how LLMs are making this possible.
How to Build a Data Stack to Win PLG, Featuring Peter Chapman
Why you should listen – To hear from a veteran data leader. Peter draws from his experience leading data teams and revenue ops to deliver some practical advice about which data initiatives you should prioritize to drive early-stage growth. He also gives some pro tips for setting up a smart data stack from the beginning.
Time Series Data Management and Data Modeling with Stanford PhD Student Tony Wang (Twitter)
Why you should listen – Because you want to nerd out on time-series data. Tune in to hear how Tony is thinking about data modeling for time series data and how he proposes bringing the tabular and time series worlds closer together.
🎥 The March Preview
Get ready to hear from these brilliant minds this month:
Kunal Agarwal – Co-founder and CEO of Unravel Data (👀 out now!)
Mike Driscoll – Co-founder at Rill Data
Kevin Liu – Co-founder and CEO of Metronome
Chad Sanderson – CEO at Gable.ai
🔗 Saved to Pocket
Databases Are Commodities. Now What? – In this one, Chris Riccomini examines a world where “every database is a PostgreSQL-compatible frontend built atop the same set of open source components,” and proposes three ways databases can still compete that are keeping the space exciting.
Labeling with Confidence – This report from the team at refuel looks at four different techniques for estimating the confidence of LLM-generated data labels and details the results.
Why data teams must separate support work from development work – Team RudderStack looks at the reason so many data teams are stuck on the ad hoc treadmill struggling to deliver competitive advantage to their businesses and proposes taking a page from the IT and software engineering to chart a path forward.
🗓 Upcoming Events
Join once and future guests of The Data Stack Show at these upcoming events:
3/13 | Cube Dev | Streamlining Data Analytics with a Semantic Layer
3/26-28 | Data Council | Data Council Austin 20% discount with promo code: datastack20
Thanks for reading! If it was worth your time, please share the newsletter with your friends, and subscribe if you haven’t yet. Oh, and we’d love to hear from you. Comment below if you have any feedback for us or just want to connect. ✌ See ya next month.
- Brooks & The Data Stack Show Team